Hugging Face's logo

KhangTruong
/
Testing-model

Image Segmentation
PyTorch
sid-unet
synthetic-image-detection
diffusion-detection
Model card Files
xet
 Community
KhangTruong commited on
Commit
788c2cf
·
verified ·
1 Parent(s): cb8a1ea

feat(release): final training report and checkpoints for KhangTruong/Testing-model (v1)

Browse files
Files changed (33) hide show
  1. .gitattributes +4 -0
  2. README.md +77 -0
  3. checkpoints/checkpoint_best.pt +3 -0
  4. checkpoints/checkpoint_best_config.yaml +120 -0
  5. checkpoints/checkpoint_latest.pt +3 -0
  6. checkpoints/checkpoint_latest_config.yaml +120 -0
  7. checkpoints/effective_config.yaml +116 -0
  8. config.yaml +120 -0
  9. manifest.json +20 -0
  10. reports/test_evaluation_report.json +212 -0
  11. reports/test_evaluation_report.md +51 -0
  12. reports/training_curves.jpg +3 -0
  13. reports/training_curves.pdf +0 -0
  14. reports/training_curves.png +3 -0
  15. reports/training_final_report.json +261 -0
  16. reports/training_final_report.md +61 -0
  17. reports/training_history.csv +2 -0
  18. reports/training_history.json +48 -0
  19. versions/v1/checkpoint_best_config.yaml +120 -0
  20. versions/v1/checkpoint_latest_config.yaml +120 -0
  21. versions/v1/checkpoints/checkpoint_best.pt +3 -0
  22. versions/v1/checkpoints/checkpoint_latest.pt +3 -0
  23. versions/v1/effective_config.yaml +116 -0
  24. versions/v1/manifest.json +20 -0
  25. versions/v1/reports/test_evaluation_report.json +212 -0
  26. versions/v1/reports/test_evaluation_report.md +51 -0
  27. versions/v1/reports/training_curves.jpg +3 -0
  28. versions/v1/reports/training_curves.pdf +0 -0
  29. versions/v1/reports/training_curves.png +3 -0
  30. versions/v1/reports/training_final_report.json +261 -0
  31. versions/v1/reports/training_final_report.md +61 -0
  32. versions/v1/reports/training_history.csv +2 -0
  33. versions/v1/reports/training_history.json +48 -0
.gitattributes CHANGED
@@ -33,3 +33,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
33
  *.zip filter=lfs diff=lfs merge=lfs -text
34
  *.zst filter=lfs diff=lfs merge=lfs -text
35
  *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ versions/v1/reports/training_curves.png filter=lfs diff=lfs merge=lfs -text
37
+ reports/training_curves.png filter=lfs diff=lfs merge=lfs -text
38
+ versions/v1/reports/training_curves.jpg filter=lfs diff=lfs merge=lfs -text
39
+ reports/training_curves.jpg filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - pytorch
5
+ - sid-unet
6
+ - image-segmentation
7
+ - synthetic-image-detection
8
+ - diffusion-detection
9
+ pipeline_tag: image-segmentation
10
+ library_name: sid-unet
11
+ ---
12
+
13
+ # Testing-model
14
+
15
+ Official repository for **Testing-model** checkpoints, evaluated on AI-generated image detection and segmentation tasks.
16
+
17
+ ## 📌 Available Versions
18
+
19
+ | Version | Best Metric Score | Best Epoch | Updated |
20
+ | :--- | :--- | :--- | :--- |
21
+ | `v1` | 0.6 | 1 | 2026-10-06 |
22
+
23
+ - **Active Version:** `v1`
24
+ - **Best Validation Score:** `0.6` (Epoch `1`)
25
+ - **Total Training Epochs:** `1`
26
+
27
+ ---
28
+
29
+ ## 🚀 Quickstart & Checkpoint Usage
30
+
31
+ ### 1. Pull checkpoints using `sid-pull`
32
+ To download the latest checkpoint directly to your local workspace:
33
+ ```bash
34
+ sid-pull --name KhangTruong/Testing-model
35
+ ```
36
+
37
+ To pull a specific version:
38
+ ```bash
39
+ sid-pull --name KhangTruong/Testing-model --version v1
40
+ ```
41
+
42
+ ### 2. Resume or Fine-tune with `sid-train`
43
+ Once downloaded, resume training seamlessly:
44
+ ```bash
45
+ sid-train --config configs/experiments/diffusion_diff_minimized/default.yaml --override training.epochs=40
46
+ ```
47
+ Or specify explicit checkpoint path:
48
+ ```bash
49
+ sid-train --resume checkpoints/checkpoint_latest.pt
50
+ ```
51
+
52
+ ### 3. Push new checkpoints using `sid-push`
53
+ ```bash
54
+ sid-push --name KhangTruong/Testing-model --version v2
55
+ ```
56
+
57
+ ---
58
+
59
+ ## 📂 Repository Layout
60
+
61
+ ```
62
+ KhangTruong/Testing-model/
63
+ ├── README.md # Model Card & Documentation
64
+ ├── config.yaml # Model & Training Configuration
65
+ ├── manifest.json # Catalog of all versions & metrics
66
+ ├── checkpoints/ # Active / Latest Checkpoints
67
+ │ ├── checkpoint_best.pt
68
+ │ ├── checkpoint_latest.pt
69
+ │ └── checkpoint_periodic.pt
70
+ ├── reports/ # Evaluation reports & plots
71
+ │ ├── training_curves.png
72
+ │ └── test_evaluation_report.json
73
+ └── versions/ # Version Archive
74
+ └── v1/
75
+ ├── checkpoints/
76
+ └── reports/
77
+ ```
checkpoints/checkpoint_best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9172bf7d3e6103e1b5a3f72fafb499bd31cef50412504f57501c9bb3bae43a4
3
+ size 217443
checkpoints/checkpoint_best_config.yaml ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ base_batch_size: 16
33
+ num_gpus: 1
34
+ _batch_size_scaled: true
35
+ model:
36
+ name: unet
37
+ in_channels: 3
38
+ out_channels: 1
39
+ features:
40
+ - 8
41
+ - 16
42
+ bilinear: true
43
+ dropout: 0.0
44
+ aux_classifier: false
45
+ num_classes: 3
46
+ gradient_checkpointing: false
47
+ loss:
48
+ mask_loss_type: combined
49
+ bce_weight: 0.5
50
+ dice_weight: 0.5
51
+ focal_weight: 0.5
52
+ focal_gamma: 2.0
53
+ focal_alpha: 0.25
54
+ aux_loss_type: cross_entropy
55
+ aux_weight: 0.2
56
+ training:
57
+ epochs: 1
58
+ learning_rate: 0.001
59
+ weight_decay: 0.0001
60
+ optimizer: adamw
61
+ use_8bit_optimizer: false
62
+ check_8bit_compatibility: true
63
+ fallback_to_16bit: true
64
+ fallback_on_unsupported_8bit: true
65
+ scheduler: cosine
66
+ resume_lr_mode: auto
67
+ resume_lr: null
68
+ warmup_epochs: 1
69
+ min_lr: 1.0e-06
70
+ grad_clip_norm: 1.0
71
+ amp: false
72
+ gradient_accumulation_steps: 1
73
+ data_parallel: false
74
+ use_hard_mining: false
75
+ auto_batch_size: true
76
+ empty_cache_per_epoch: true
77
+ save_best: true
78
+ save_latest: true
79
+ checkpoint_period: 3600
80
+ eval_interval: 1
81
+ early_stopping_patience: 2
82
+ early_stopping_metric: val_iou
83
+ early_stopping_mode: max
84
+ push_to_hub: true
85
+ hub_repo: KhangTruong/Testing-model
86
+ hub_version: v1
87
+ batch_size: 2
88
+ hf_repo: KhangTruong/Testing-model
89
+ num_gpus: 1
90
+ hard_mining:
91
+ enabled: false
92
+ metric: median
93
+ reset_epochs: 5
94
+ bootstrapping:
95
+ enabled: false
96
+ run_bootstrap: false
97
+ epochs: 5
98
+ num_samples: 512
99
+ batch_size: null
100
+ learning_rate: null
101
+ freeze_strategy: channel_stream
102
+ stream_ratio: 0.5
103
+ release_mode: restore
104
+ freeze_modules: []
105
+ initialization: kaiming_normal
106
+ target_score: 0.5
107
+ min_loss_drop: 0.15
108
+ early_stopping: false
109
+ patience: 3
110
+ logging:
111
+ log_interval: 1
112
+ log_memory: true
113
+ measure_network: true
114
+ post_processing:
115
+ enabled: true
116
+ min_area: 64
117
+ fill_holes: true
118
+ max_hole_size: 256
119
+ morphology: open_close
120
+ morph_kernel_size: 3
checkpoints/checkpoint_latest.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a40f37c060ab46b418b0e7ae2ec1c17db6b65f29ff9b38bfcb64630e174cdbd3
3
+ size 217771
checkpoints/checkpoint_latest_config.yaml ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ base_batch_size: 16
33
+ num_gpus: 1
34
+ _batch_size_scaled: true
35
+ model:
36
+ name: unet
37
+ in_channels: 3
38
+ out_channels: 1
39
+ features:
40
+ - 8
41
+ - 16
42
+ bilinear: true
43
+ dropout: 0.0
44
+ aux_classifier: false
45
+ num_classes: 3
46
+ gradient_checkpointing: false
47
+ loss:
48
+ mask_loss_type: combined
49
+ bce_weight: 0.5
50
+ dice_weight: 0.5
51
+ focal_weight: 0.5
52
+ focal_gamma: 2.0
53
+ focal_alpha: 0.25
54
+ aux_loss_type: cross_entropy
55
+ aux_weight: 0.2
56
+ training:
57
+ epochs: 1
58
+ learning_rate: 0.001
59
+ weight_decay: 0.0001
60
+ optimizer: adamw
61
+ use_8bit_optimizer: false
62
+ check_8bit_compatibility: true
63
+ fallback_to_16bit: true
64
+ fallback_on_unsupported_8bit: true
65
+ scheduler: cosine
66
+ resume_lr_mode: auto
67
+ resume_lr: null
68
+ warmup_epochs: 1
69
+ min_lr: 1.0e-06
70
+ grad_clip_norm: 1.0
71
+ amp: false
72
+ gradient_accumulation_steps: 1
73
+ data_parallel: false
74
+ use_hard_mining: false
75
+ auto_batch_size: true
76
+ empty_cache_per_epoch: true
77
+ save_best: true
78
+ save_latest: true
79
+ checkpoint_period: 3600
80
+ eval_interval: 1
81
+ early_stopping_patience: 2
82
+ early_stopping_metric: val_iou
83
+ early_stopping_mode: max
84
+ push_to_hub: true
85
+ hub_repo: KhangTruong/Testing-model
86
+ hub_version: v1
87
+ batch_size: 2
88
+ hf_repo: KhangTruong/Testing-model
89
+ num_gpus: 1
90
+ hard_mining:
91
+ enabled: false
92
+ metric: median
93
+ reset_epochs: 5
94
+ bootstrapping:
95
+ enabled: false
96
+ run_bootstrap: false
97
+ epochs: 5
98
+ num_samples: 512
99
+ batch_size: null
100
+ learning_rate: null
101
+ freeze_strategy: channel_stream
102
+ stream_ratio: 0.5
103
+ release_mode: restore
104
+ freeze_modules: []
105
+ initialization: kaiming_normal
106
+ target_score: 0.5
107
+ min_loss_drop: 0.15
108
+ early_stopping: false
109
+ patience: 3
110
+ logging:
111
+ log_interval: 1
112
+ log_memory: true
113
+ measure_network: true
114
+ post_processing:
115
+ enabled: true
116
+ min_area: 64
117
+ fill_holes: true
118
+ max_hole_size: 256
119
+ morphology: open_close
120
+ morph_kernel_size: 3
checkpoints/effective_config.yaml ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ model:
33
+ name: unet
34
+ in_channels: 3
35
+ out_channels: 1
36
+ features:
37
+ - 8
38
+ - 16
39
+ bilinear: true
40
+ dropout: 0.0
41
+ aux_classifier: false
42
+ num_classes: 3
43
+ gradient_checkpointing: false
44
+ loss:
45
+ mask_loss_type: combined
46
+ bce_weight: 0.5
47
+ dice_weight: 0.5
48
+ focal_weight: 0.5
49
+ focal_gamma: 2.0
50
+ focal_alpha: 0.25
51
+ aux_loss_type: cross_entropy
52
+ aux_weight: 0.2
53
+ training:
54
+ epochs: 1
55
+ learning_rate: 0.001
56
+ weight_decay: 0.0001
57
+ optimizer: adamw
58
+ use_8bit_optimizer: false
59
+ check_8bit_compatibility: true
60
+ fallback_to_16bit: true
61
+ fallback_on_unsupported_8bit: true
62
+ scheduler: cosine
63
+ resume_lr_mode: auto
64
+ resume_lr: null
65
+ warmup_epochs: 1
66
+ min_lr: 1.0e-06
67
+ grad_clip_norm: 1.0
68
+ amp: false
69
+ gradient_accumulation_steps: 1
70
+ data_parallel: true
71
+ use_hard_mining: false
72
+ auto_batch_size: true
73
+ empty_cache_per_epoch: true
74
+ save_best: true
75
+ save_latest: true
76
+ checkpoint_period: 3600
77
+ eval_interval: 1
78
+ early_stopping_patience: 2
79
+ early_stopping_metric: val_iou
80
+ early_stopping_mode: max
81
+ push_to_hub: true
82
+ hub_repo: KhangTruong/Testing-model
83
+ hub_version: v1
84
+ batch_size: 2
85
+ hf_repo: KhangTruong/Testing-model
86
+ hard_mining:
87
+ enabled: false
88
+ metric: median
89
+ reset_epochs: 5
90
+ bootstrapping:
91
+ enabled: false
92
+ run_bootstrap: false
93
+ epochs: 5
94
+ num_samples: 512
95
+ batch_size: null
96
+ learning_rate: null
97
+ freeze_strategy: channel_stream
98
+ stream_ratio: 0.5
99
+ release_mode: restore
100
+ freeze_modules: []
101
+ initialization: kaiming_normal
102
+ target_score: 0.5
103
+ min_loss_drop: 0.15
104
+ early_stopping: false
105
+ patience: 3
106
+ logging:
107
+ log_interval: 1
108
+ log_memory: true
109
+ measure_network: true
110
+ post_processing:
111
+ enabled: true
112
+ min_area: 64
113
+ fill_holes: true
114
+ max_hole_size: 256
115
+ morphology: open_close
116
+ morph_kernel_size: 3
config.yaml ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ base_batch_size: 16
33
+ num_gpus: 1
34
+ _batch_size_scaled: true
35
+ model:
36
+ name: unet
37
+ in_channels: 3
38
+ out_channels: 1
39
+ features:
40
+ - 8
41
+ - 16
42
+ bilinear: true
43
+ dropout: 0.0
44
+ aux_classifier: false
45
+ num_classes: 3
46
+ gradient_checkpointing: false
47
+ loss:
48
+ mask_loss_type: combined
49
+ bce_weight: 0.5
50
+ dice_weight: 0.5
51
+ focal_weight: 0.5
52
+ focal_gamma: 2.0
53
+ focal_alpha: 0.25
54
+ aux_loss_type: cross_entropy
55
+ aux_weight: 0.2
56
+ training:
57
+ epochs: 1
58
+ learning_rate: 0.001
59
+ weight_decay: 0.0001
60
+ optimizer: adamw
61
+ use_8bit_optimizer: false
62
+ check_8bit_compatibility: true
63
+ fallback_to_16bit: true
64
+ fallback_on_unsupported_8bit: true
65
+ scheduler: cosine
66
+ resume_lr_mode: auto
67
+ resume_lr: null
68
+ warmup_epochs: 1
69
+ min_lr: 1.0e-06
70
+ grad_clip_norm: 1.0
71
+ amp: false
72
+ gradient_accumulation_steps: 1
73
+ data_parallel: false
74
+ use_hard_mining: false
75
+ auto_batch_size: true
76
+ empty_cache_per_epoch: true
77
+ save_best: true
78
+ save_latest: true
79
+ checkpoint_period: 3600
80
+ eval_interval: 1
81
+ early_stopping_patience: 2
82
+ early_stopping_metric: val_iou
83
+ early_stopping_mode: max
84
+ push_to_hub: true
85
+ hub_repo: KhangTruong/Testing-model
86
+ hub_version: v1
87
+ batch_size: 2
88
+ hf_repo: KhangTruong/Testing-model
89
+ num_gpus: 1
90
+ hard_mining:
91
+ enabled: false
92
+ metric: median
93
+ reset_epochs: 5
94
+ bootstrapping:
95
+ enabled: false
96
+ run_bootstrap: false
97
+ epochs: 5
98
+ num_samples: 512
99
+ batch_size: null
100
+ learning_rate: null
101
+ freeze_strategy: channel_stream
102
+ stream_ratio: 0.5
103
+ release_mode: restore
104
+ freeze_modules: []
105
+ initialization: kaiming_normal
106
+ target_score: 0.5
107
+ min_loss_drop: 0.15
108
+ early_stopping: false
109
+ patience: 3
110
+ logging:
111
+ log_interval: 1
112
+ log_memory: true
113
+ measure_network: true
114
+ post_processing:
115
+ enabled: true
116
+ min_area: 64
117
+ fill_holes: true
118
+ max_hole_size: 256
119
+ morphology: open_close
120
+ morph_kernel_size: 3
manifest.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "versions": [
3
+ "v1"
4
+ ],
5
+ "versions_meta": {
6
+ "v1": {
7
+ "version": "v1",
8
+ "best_score": "0.6",
9
+ "best_epoch": "1",
10
+ "total_epochs": "1",
11
+ "step": "1",
12
+ "updated_at": "2026-10-06T12:25:39.977544+00:00",
13
+ "files": [
14
+ "checkpoint_best.pt",
15
+ "checkpoint_latest.pt"
16
+ ]
17
+ }
18
+ },
19
+ "current_version": "v1"
20
+ }
reports/test_evaluation_report.json ADDED
@@ -0,0 +1,212 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "overall_metrics": {
3
+ "test_bce_loss": 0.6941993832588196,
4
+ "test_dice_loss": 0.6576691269874573,
5
+ "test_focal_loss": 0.09276691824197769,
6
+ "test_mask_loss": 0.7223176956176758,
7
+ "test_total_loss": 0.7223176956176758,
8
+ "test_iou": 0.5,
9
+ "test_dice": 0.5,
10
+ "test_f1": 0.5,
11
+ "test_pixel_f1": 0.5,
12
+ "test_auroc": 0.7407521173590794,
13
+ "test_pixel_auroc": 0.7407521173590794,
14
+ "test_pixel_acc": 0.5,
15
+ "test_precision": 0.5,
16
+ "test_recall": 0.5,
17
+ "test_specificity": 0.5
18
+ },
19
+ "per_label_metrics": {
20
+ "Label 0 (Real / Black Mask)": {
21
+ "iou": 0.0,
22
+ "dice": 0.0,
23
+ "f1": 0.0,
24
+ "pixel_f1": 0.0,
25
+ "auroc": 0.48150423471815884,
26
+ "pixel_auroc": 0.48150423471815884,
27
+ "pixel_acc": 0.0,
28
+ "precision": 0.0,
29
+ "recall": 0.0,
30
+ "specificity": 0.0,
31
+ "samples": 1
32
+ },
33
+ "Label 1 (Synthetic / White Mask)": {
34
+ "iou": 1.0,
35
+ "dice": 1.0,
36
+ "f1": 1.0,
37
+ "pixel_f1": 1.0,
38
+ "auroc": 1.0,
39
+ "pixel_auroc": 1.0,
40
+ "pixel_acc": 1.0,
41
+ "precision": 1.0,
42
+ "recall": 1.0,
43
+ "specificity": 1.0,
44
+ "samples": 1
45
+ },
46
+ "Label 2 (Tampered / Partial Mask)": {
47
+ "iou": 0.0,
48
+ "dice": 0.0,
49
+ "f1": 0.0,
50
+ "pixel_f1": 0.0,
51
+ "auroc": 0.0,
52
+ "pixel_auroc": 0.0,
53
+ "pixel_acc": 0.0,
54
+ "precision": 0.0,
55
+ "recall": 0.0,
56
+ "specificity": 0.0,
57
+ "samples": 0
58
+ }
59
+ },
60
+ "configuration": {
61
+ "project": {
62
+ "name": "test_smoke",
63
+ "seed": 42,
64
+ "device": "cpu",
65
+ "output_dir": "/tmp/tmp08k8509s/out/RUN/test_smoke",
66
+ "hub_repo": null,
67
+ "resume_repo": null
68
+ },
69
+ "data": {
70
+ "dataset_name": "KhangTruong/IMD2020",
71
+ "streaming": false,
72
+ "image_size": [
73
+ 32,
74
+ 32
75
+ ],
76
+ "batch_size": 16,
77
+ "num_workers": 0,
78
+ "pin_memory": false,
79
+ "shuffle_buffer_size": 1000,
80
+ "train_split": "train",
81
+ "val_split": "validation",
82
+ "test_split": "test",
83
+ "eval_split": "test",
84
+ "train_samples_per_epoch": 2,
85
+ "val_samples_per_epoch": -1,
86
+ "val_samples": 2,
87
+ "test_samples": 2,
88
+ "evaluate_on_test": true,
89
+ "augmentations": {
90
+ "horizontal_flip": 0.5,
91
+ "vertical_flip": 0.0,
92
+ "random_rotate90": 0.0,
93
+ "color_jitter": 0.0
94
+ },
95
+ "base_batch_size": 16,
96
+ "num_gpus": 1,
97
+ "_batch_size_scaled": true
98
+ },
99
+ "model": {
100
+ "name": "unet",
101
+ "in_channels": 3,
102
+ "out_channels": 1,
103
+ "features": [
104
+ 8,
105
+ 16
106
+ ],
107
+ "bilinear": true,
108
+ "dropout": 0.0,
109
+ "aux_classifier": false,
110
+ "num_classes": 3,
111
+ "gradient_checkpointing": false
112
+ },
113
+ "loss": {
114
+ "mask_loss_type": "combined",
115
+ "bce_weight": 0.5,
116
+ "dice_weight": 0.5,
117
+ "focal_weight": 0.5,
118
+ "focal_gamma": 2.0,
119
+ "focal_alpha": 0.25,
120
+ "aux_loss_type": "cross_entropy",
121
+ "aux_weight": 0.2
122
+ },
123
+ "training": {
124
+ "epochs": 1,
125
+ "learning_rate": 0.001,
126
+ "weight_decay": 0.0001,
127
+ "optimizer": "adamw",
128
+ "use_8bit_optimizer": false,
129
+ "check_8bit_compatibility": true,
130
+ "fallback_to_16bit": true,
131
+ "fallback_on_unsupported_8bit": true,
132
+ "scheduler": "cosine",
133
+ "resume_lr_mode": "auto",
134
+ "resume_lr": null,
135
+ "warmup_epochs": 1,
136
+ "min_lr": 1e-06,
137
+ "grad_clip_norm": 1.0,
138
+ "amp": false,
139
+ "gradient_accumulation_steps": 1,
140
+ "data_parallel": false,
141
+ "use_hard_mining": false,
142
+ "auto_batch_size": true,
143
+ "empty_cache_per_epoch": true,
144
+ "save_best": true,
145
+ "save_latest": true,
146
+ "checkpoint_period": 3600,
147
+ "eval_interval": 1,
148
+ "early_stopping_patience": 2,
149
+ "early_stopping_metric": "val_iou",
150
+ "early_stopping_mode": "max",
151
+ "push_to_hub": true,
152
+ "hub_repo": "KhangTruong/Testing-model",
153
+ "hub_version": "v1",
154
+ "batch_size": 2,
155
+ "hf_repo": "KhangTruong/Testing-model",
156
+ "num_gpus": 1
157
+ },
158
+ "hard_mining": {
159
+ "enabled": false,
160
+ "metric": "median",
161
+ "reset_epochs": 5
162
+ },
163
+ "bootstrapping": {
164
+ "enabled": false,
165
+ "run_bootstrap": false,
166
+ "epochs": 5,
167
+ "num_samples": 512,
168
+ "batch_size": null,
169
+ "learning_rate": null,
170
+ "freeze_strategy": "channel_stream",
171
+ "stream_ratio": 0.5,
172
+ "release_mode": "restore",
173
+ "freeze_modules": [],
174
+ "initialization": "kaiming_normal",
175
+ "target_score": 0.5,
176
+ "min_loss_drop": 0.15,
177
+ "early_stopping": false,
178
+ "patience": 3
179
+ },
180
+ "logging": {
181
+ "log_interval": 1,
182
+ "log_memory": true,
183
+ "measure_network": true
184
+ },
185
+ "post_processing": {
186
+ "enabled": true,
187
+ "min_area": 64,
188
+ "fill_holes": true,
189
+ "max_hole_size": 256,
190
+ "morphology": "open_close",
191
+ "morph_kernel_size": 3
192
+ }
193
+ },
194
+ "hyperparameters": {
195
+ "Project Name": "test_smoke",
196
+ "Dataset": "KhangTruong/IMD2020",
197
+ "Splits": "Train: 'train' | Val: 'validation' | Test: 'test'",
198
+ "Streaming": false,
199
+ "Image Size": "[32, 32]",
200
+ "Batch Size": 16,
201
+ "Epoch Limit / Steps": "2 samples/epoch",
202
+ "Model": "unet (features=[8, 16], bilinear=True)",
203
+ "Aux Classifier": "False (classes=3)",
204
+ "Mask Loss": "Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5)",
205
+ "Aux Loss": "cross_entropy (wt=0.2)",
206
+ "Post-Processing": "Enabled (min_area=64, fill_holes=True, morph=open_close)",
207
+ "Optimizer": "adamw (lr=0.001, wd=0.0001)",
208
+ "Scheduler": "cosine",
209
+ "Epochs": 1,
210
+ "AMP": false
211
+ }
212
+ }
reports/test_evaluation_report.md ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SID-UNet Evaluation Report
2
+
3
+ ### Run Configuration & Hyperparameters
4
+
5
+ | Hyperparameter / Setting | Configured Value |
6
+ |----------------------------|----------------------------------------------------------|
7
+ | Project Name | test_smoke |
8
+ | Dataset | KhangTruong/IMD2020 |
9
+ | Splits | Train: 'train' | Val: 'validation' | Test: 'test' |
10
+ | Streaming | False |
11
+ | Image Size | [32, 32] |
12
+ | Batch Size | 16 |
13
+ | Epoch Limit / Steps | 2 samples/epoch |
14
+ | Model | unet (features=[8, 16], bilinear=True) |
15
+ | Aux Classifier | False (classes=3) |
16
+ | Mask Loss | Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5) |
17
+ | Aux Loss | cross_entropy (wt=0.2) |
18
+ | Post-Processing | Enabled (min_area=64, fill_holes=True, morph=open_close) |
19
+ | Optimizer | adamw (lr=0.001, wd=0.0001) |
20
+ | Scheduler | cosine |
21
+ | Epochs | 1 |
22
+ | AMP | False |
23
+
24
+ ### Overall Segmentation & Classification Metrics
25
+
26
+ | Metric | Value |
27
+ |------------------|---------|
28
+ | Test Bce Loss | 0.6942 |
29
+ | Test Dice Loss | 0.6577 |
30
+ | Test Focal Loss | 0.0928 |
31
+ | Test Mask Loss | 0.7223 |
32
+ | Test Total Loss | 0.7223 |
33
+ | Test Iou | 0.5 |
34
+ | Test Dice | 0.5 |
35
+ | Test F1 | 0.5 |
36
+ | Test Pixel F1 | 0.5 |
37
+ | Test Auroc | 0.7408 |
38
+ | Test Pixel Auroc | 0.7408 |
39
+ | Test Pixel Acc | 0.5 |
40
+ | Test Precision | 0.5 |
41
+ | Test Recall | 0.5 |
42
+ | Test Specificity | 0.5 |
43
+
44
+ ### Per-Subset Metrics Breakdown
45
+
46
+ | Subset | Mean IoU | Dice / F1 | Pixel Accuracy | Sample Count |
47
+ |-----------------------------------|------------|-------------|------------------|----------------|
48
+ | Label 0 (Real / Black Mask) | 0 | 0 | 0 | 1 |
49
+ | Label 1 (Synthetic / White Mask) | 1 | 1 | 1 | 1 |
50
+ | Label 2 (Tampered / Partial Mask) | 0 | 0 | 0 | 0 |
51
+
reports/training_curves.jpg ADDED

Git LFS Details

  • SHA256: 71a598c9822d6cd64d2fbcf319549f66c7b0b6a6ec810076fe0563f50e1779ae
  • Pointer size: 131 Bytes
  • Size of remote file: 512 kB
reports/training_curves.pdf ADDED
Binary file (38.3 kB). View file
 
reports/training_curves.png ADDED

Git LFS Details

  • SHA256: fc236cd58e55a20de123a51c994b7cdcfd690e1dedb6c41e9abafee038fd6a70
  • Pointer size: 131 Bytes
  • Size of remote file: 362 kB
reports/training_final_report.json ADDED
@@ -0,0 +1,261 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "overall_metrics": {
3
+ "val_bce_loss": 0.6869940757751465,
4
+ "val_dice_loss": 0.590221107006073,
5
+ "val_focal_loss": 0.08142829686403275,
6
+ "val_mask_loss": 0.6793217658996582,
7
+ "val_total_loss": 0.6793217658996582,
8
+ "val_iou": 0.6,
9
+ "val_dice": 0.6,
10
+ "val_f1": 0.6,
11
+ "val_pixel_f1": 0.6,
12
+ "val_auroc": 0.79323464758927,
13
+ "val_pixel_auroc": 0.79323464758927,
14
+ "val_pixel_acc": 0.6,
15
+ "val_precision": 0.6,
16
+ "val_recall": 0.6,
17
+ "val_specificity": 0.6
18
+ },
19
+ "per_label_metrics": {
20
+ "Label 0 (Real / Black Mask)": {
21
+ "iou": 0.0,
22
+ "dice": 0.0,
23
+ "f1": 0.0,
24
+ "pixel_f1": 0.0,
25
+ "auroc": 0.48308661897317506,
26
+ "pixel_auroc": 0.48308661897317506,
27
+ "pixel_acc": 0.0,
28
+ "precision": 0.0,
29
+ "recall": 0.0,
30
+ "specificity": 0.0,
31
+ "samples": 4
32
+ },
33
+ "Label 1 (Synthetic / White Mask)": {
34
+ "iou": 1.0,
35
+ "dice": 1.0,
36
+ "f1": 1.0,
37
+ "pixel_f1": 1.0,
38
+ "auroc": 1.0,
39
+ "pixel_auroc": 1.0,
40
+ "pixel_acc": 1.0,
41
+ "precision": 1.0,
42
+ "recall": 1.0,
43
+ "specificity": 1.0,
44
+ "samples": 3
45
+ },
46
+ "Label 2 (Tampered / Partial Mask)": {
47
+ "iou": 1.0,
48
+ "dice": 1.0,
49
+ "f1": 1.0,
50
+ "pixel_f1": 1.0,
51
+ "auroc": 1.0,
52
+ "pixel_auroc": 1.0,
53
+ "pixel_acc": 1.0,
54
+ "precision": 1.0,
55
+ "recall": 1.0,
56
+ "specificity": 1.0,
57
+ "samples": 3
58
+ }
59
+ },
60
+ "curves_plot_path": "/tmp/tmp08k8509s/out/RUN/test_smoke/reports/training_curves.png",
61
+ "training_history": [
62
+ {
63
+ "epoch": 1,
64
+ "lr": 0.001,
65
+ "train_loss": 0.7411096692085266,
66
+ "train_mask_loss": 0.7411096692085266,
67
+ "train_aux_loss": 0.0,
68
+ "train_iou": 0.17578125,
69
+ "val_loss": 0.6793217658996582,
70
+ "val_mask_loss": 0.6793217658996582,
71
+ "val_aux_loss": 0.0,
72
+ "val_iou": 0.6,
73
+ "val_dice": 0.6,
74
+ "val_f1": 0.6,
75
+ "val_pixel_f1": 0.6,
76
+ "val_auroc": 0.79323464758927,
77
+ "val_pixel_auroc": 0.79323464758927,
78
+ "val_pixel_acc": 0.6,
79
+ "val_precision": 0.6,
80
+ "val_recall": 0.6,
81
+ "val_aux_accuracy": 0.0,
82
+ "train_metrics": {
83
+ "bce_loss": 0.7112534642219543,
84
+ "dice_loss": 0.673093318939209,
85
+ "focal_loss": 0.09787261486053467,
86
+ "mask_loss": 0.7411096692085266,
87
+ "total_loss": 0.7411096692085266,
88
+ "iou": 0.17578125
89
+ },
90
+ "val_metrics": {
91
+ "val_bce_loss": 0.6869940757751465,
92
+ "val_dice_loss": 0.590221107006073,
93
+ "val_focal_loss": 0.08142829686403275,
94
+ "val_mask_loss": 0.6793217658996582,
95
+ "val_total_loss": 0.6793217658996582,
96
+ "val_iou": 0.6,
97
+ "val_dice": 0.6,
98
+ "val_f1": 0.6,
99
+ "val_pixel_f1": 0.6,
100
+ "val_auroc": 0.79323464758927,
101
+ "val_pixel_auroc": 0.79323464758927,
102
+ "val_pixel_acc": 0.6,
103
+ "val_precision": 0.6,
104
+ "val_recall": 0.6,
105
+ "val_specificity": 0.6
106
+ }
107
+ }
108
+ ],
109
+ "configuration": {
110
+ "project": {
111
+ "name": "test_smoke",
112
+ "seed": 42,
113
+ "device": "cpu",
114
+ "output_dir": "/tmp/tmp08k8509s/out/RUN/test_smoke",
115
+ "hub_repo": null,
116
+ "resume_repo": null
117
+ },
118
+ "data": {
119
+ "dataset_name": "KhangTruong/IMD2020",
120
+ "streaming": false,
121
+ "image_size": [
122
+ 32,
123
+ 32
124
+ ],
125
+ "batch_size": 16,
126
+ "num_workers": 0,
127
+ "pin_memory": false,
128
+ "shuffle_buffer_size": 1000,
129
+ "train_split": "train",
130
+ "val_split": "validation",
131
+ "test_split": "test",
132
+ "eval_split": "test",
133
+ "train_samples_per_epoch": 2,
134
+ "val_samples_per_epoch": -1,
135
+ "val_samples": 2,
136
+ "test_samples": 2,
137
+ "evaluate_on_test": true,
138
+ "augmentations": {
139
+ "horizontal_flip": 0.5,
140
+ "vertical_flip": 0.0,
141
+ "random_rotate90": 0.0,
142
+ "color_jitter": 0.0
143
+ },
144
+ "base_batch_size": 16,
145
+ "num_gpus": 1,
146
+ "_batch_size_scaled": true
147
+ },
148
+ "model": {
149
+ "name": "unet",
150
+ "in_channels": 3,
151
+ "out_channels": 1,
152
+ "features": [
153
+ 8,
154
+ 16
155
+ ],
156
+ "bilinear": true,
157
+ "dropout": 0.0,
158
+ "aux_classifier": false,
159
+ "num_classes": 3,
160
+ "gradient_checkpointing": false
161
+ },
162
+ "loss": {
163
+ "mask_loss_type": "combined",
164
+ "bce_weight": 0.5,
165
+ "dice_weight": 0.5,
166
+ "focal_weight": 0.5,
167
+ "focal_gamma": 2.0,
168
+ "focal_alpha": 0.25,
169
+ "aux_loss_type": "cross_entropy",
170
+ "aux_weight": 0.2
171
+ },
172
+ "training": {
173
+ "epochs": 1,
174
+ "learning_rate": 0.001,
175
+ "weight_decay": 0.0001,
176
+ "optimizer": "adamw",
177
+ "use_8bit_optimizer": false,
178
+ "check_8bit_compatibility": true,
179
+ "fallback_to_16bit": true,
180
+ "fallback_on_unsupported_8bit": true,
181
+ "scheduler": "cosine",
182
+ "resume_lr_mode": "auto",
183
+ "resume_lr": null,
184
+ "warmup_epochs": 1,
185
+ "min_lr": 1e-06,
186
+ "grad_clip_norm": 1.0,
187
+ "amp": false,
188
+ "gradient_accumulation_steps": 1,
189
+ "data_parallel": false,
190
+ "use_hard_mining": false,
191
+ "auto_batch_size": true,
192
+ "empty_cache_per_epoch": true,
193
+ "save_best": true,
194
+ "save_latest": true,
195
+ "checkpoint_period": 3600,
196
+ "eval_interval": 1,
197
+ "early_stopping_patience": 2,
198
+ "early_stopping_metric": "val_iou",
199
+ "early_stopping_mode": "max",
200
+ "push_to_hub": true,
201
+ "hub_repo": "KhangTruong/Testing-model",
202
+ "hub_version": "v1",
203
+ "batch_size": 2,
204
+ "hf_repo": "KhangTruong/Testing-model",
205
+ "num_gpus": 1
206
+ },
207
+ "hard_mining": {
208
+ "enabled": false,
209
+ "metric": "median",
210
+ "reset_epochs": 5
211
+ },
212
+ "bootstrapping": {
213
+ "enabled": false,
214
+ "run_bootstrap": false,
215
+ "epochs": 5,
216
+ "num_samples": 512,
217
+ "batch_size": null,
218
+ "learning_rate": null,
219
+ "freeze_strategy": "channel_stream",
220
+ "stream_ratio": 0.5,
221
+ "release_mode": "restore",
222
+ "freeze_modules": [],
223
+ "initialization": "kaiming_normal",
224
+ "target_score": 0.5,
225
+ "min_loss_drop": 0.15,
226
+ "early_stopping": false,
227
+ "patience": 3
228
+ },
229
+ "logging": {
230
+ "log_interval": 1,
231
+ "log_memory": true,
232
+ "measure_network": true
233
+ },
234
+ "post_processing": {
235
+ "enabled": true,
236
+ "min_area": 64,
237
+ "fill_holes": true,
238
+ "max_hole_size": 256,
239
+ "morphology": "open_close",
240
+ "morph_kernel_size": 3
241
+ }
242
+ },
243
+ "hyperparameters": {
244
+ "Project Name": "test_smoke",
245
+ "Dataset": "KhangTruong/IMD2020",
246
+ "Splits": "Train: 'train' | Val: 'validation' | Test: 'test'",
247
+ "Streaming": false,
248
+ "Image Size": "[32, 32]",
249
+ "Batch Size": 16,
250
+ "Epoch Limit / Steps": "2 samples/epoch",
251
+ "Model": "unet (features=[8, 16], bilinear=True)",
252
+ "Aux Classifier": "False (classes=3)",
253
+ "Mask Loss": "Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5)",
254
+ "Aux Loss": "cross_entropy (wt=0.2)",
255
+ "Post-Processing": "Enabled (min_area=64, fill_holes=True, morph=open_close)",
256
+ "Optimizer": "adamw (lr=0.001, wd=0.0001)",
257
+ "Scheduler": "cosine",
258
+ "Epochs": 1,
259
+ "AMP": false
260
+ }
261
+ }
reports/training_final_report.md ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SID-UNet Evaluation Report
2
+
3
+ ### Run Configuration & Hyperparameters
4
+
5
+ | Hyperparameter / Setting | Configured Value |
6
+ |----------------------------|----------------------------------------------------------|
7
+ | Project Name | test_smoke |
8
+ | Dataset | KhangTruong/IMD2020 |
9
+ | Splits | Train: 'train' | Val: 'validation' | Test: 'test' |
10
+ | Streaming | False |
11
+ | Image Size | [32, 32] |
12
+ | Batch Size | 16 |
13
+ | Epoch Limit / Steps | 2 samples/epoch |
14
+ | Model | unet (features=[8, 16], bilinear=True) |
15
+ | Aux Classifier | False (classes=3) |
16
+ | Mask Loss | Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5) |
17
+ | Aux Loss | cross_entropy (wt=0.2) |
18
+ | Post-Processing | Enabled (min_area=64, fill_holes=True, morph=open_close) |
19
+ | Optimizer | adamw (lr=0.001, wd=0.0001) |
20
+ | Scheduler | cosine |
21
+ | Epochs | 1 |
22
+ | AMP | False |
23
+
24
+ ### Overall Segmentation & Classification Metrics
25
+
26
+ | Metric | Value |
27
+ |-----------------|---------|
28
+ | Val Bce Loss | 0.687 |
29
+ | Val Dice Loss | 0.5902 |
30
+ | Val Focal Loss | 0.0814 |
31
+ | Val Mask Loss | 0.6793 |
32
+ | Val Total Loss | 0.6793 |
33
+ | Val Iou | 0.6 |
34
+ | Val Dice | 0.6 |
35
+ | Val F1 | 0.6 |
36
+ | Val Pixel F1 | 0.6 |
37
+ | Val Auroc | 0.7932 |
38
+ | Val Pixel Auroc | 0.7932 |
39
+ | Val Pixel Acc | 0.6 |
40
+ | Val Precision | 0.6 |
41
+ | Val Recall | 0.6 |
42
+ | Val Specificity | 0.6 |
43
+
44
+ ### Training & Validation Curves
45
+
46
+ ![Training Curves](training_curves.png)
47
+
48
+ ### Epoch-by-Epoch Training & Validation Progression
49
+
50
+ | Epoch | Train Loss | Val Loss | Val IoU | Val F1 | Val AUROC | Val Pixel Acc | Learning Rate |
51
+ |---------|--------------|------------|-----------|----------|-------------|-----------------|-----------------|
52
+ | 1 | 0.7411 | 0.6793 | 0.6 | 0.6 | 0.7932 | 0.6 | 0.001 |
53
+
54
+ ### Per-Subset Metrics Breakdown
55
+
56
+ | Subset | Mean IoU | Dice / F1 | Pixel Accuracy | Sample Count |
57
+ |-----------------------------------|------------|-------------|------------------|----------------|
58
+ | Label 0 (Real / Black Mask) | 0 | 0 | 0 | 4 |
59
+ | Label 1 (Synthetic / White Mask) | 1 | 1 | 1 | 3 |
60
+ | Label 2 (Tampered / Partial Mask) | 1 | 1 | 1 | 3 |
61
+
reports/training_history.csv ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ epoch,lr,train_aux_loss,train_iou,train_loss,train_mask_loss,train_metrics_bce_loss,train_metrics_dice_loss,train_metrics_focal_loss,train_metrics_iou,train_metrics_mask_loss,train_metrics_total_loss,val_auroc,val_aux_accuracy,val_aux_loss,val_dice,val_f1,val_iou,val_loss,val_mask_loss,val_metrics_val_auroc,val_metrics_val_bce_loss,val_metrics_val_dice,val_metrics_val_dice_loss,val_metrics_val_f1,val_metrics_val_focal_loss,val_metrics_val_iou,val_metrics_val_mask_loss,val_metrics_val_pixel_acc,val_metrics_val_pixel_auroc,val_metrics_val_pixel_f1,val_metrics_val_precision,val_metrics_val_recall,val_metrics_val_specificity,val_metrics_val_total_loss,val_pixel_acc,val_pixel_auroc,val_pixel_f1,val_precision,val_recall
2
+ 1,0.001,0.0,0.17578125,0.7411096692085266,0.7411096692085266,0.7112534642219543,0.673093318939209,0.09787261486053467,0.17578125,0.7411096692085266,0.7411096692085266,0.79323464758927,0.0,0.0,0.6,0.6,0.6,0.6793217658996582,0.6793217658996582,0.79323464758927,0.6869940757751465,0.6,0.590221107006073,0.6,0.08142829686403275,0.6,0.6793217658996582,0.6,0.79323464758927,0.6,0.6,0.6,0.6,0.6793217658996582,0.6,0.79323464758927,0.6,0.6,0.6
reports/training_history.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "epoch": 1,
4
+ "lr": 0.001,
5
+ "train_loss": 0.7411096692085266,
6
+ "train_mask_loss": 0.7411096692085266,
7
+ "train_aux_loss": 0.0,
8
+ "train_iou": 0.17578125,
9
+ "val_loss": 0.6793217658996582,
10
+ "val_mask_loss": 0.6793217658996582,
11
+ "val_aux_loss": 0.0,
12
+ "val_iou": 0.6,
13
+ "val_dice": 0.6,
14
+ "val_f1": 0.6,
15
+ "val_pixel_f1": 0.6,
16
+ "val_auroc": 0.79323464758927,
17
+ "val_pixel_auroc": 0.79323464758927,
18
+ "val_pixel_acc": 0.6,
19
+ "val_precision": 0.6,
20
+ "val_recall": 0.6,
21
+ "val_aux_accuracy": 0.0,
22
+ "train_metrics": {
23
+ "bce_loss": 0.7112534642219543,
24
+ "dice_loss": 0.673093318939209,
25
+ "focal_loss": 0.09787261486053467,
26
+ "mask_loss": 0.7411096692085266,
27
+ "total_loss": 0.7411096692085266,
28
+ "iou": 0.17578125
29
+ },
30
+ "val_metrics": {
31
+ "val_bce_loss": 0.6869940757751465,
32
+ "val_dice_loss": 0.590221107006073,
33
+ "val_focal_loss": 0.08142829686403275,
34
+ "val_mask_loss": 0.6793217658996582,
35
+ "val_total_loss": 0.6793217658996582,
36
+ "val_iou": 0.6,
37
+ "val_dice": 0.6,
38
+ "val_f1": 0.6,
39
+ "val_pixel_f1": 0.6,
40
+ "val_auroc": 0.79323464758927,
41
+ "val_pixel_auroc": 0.79323464758927,
42
+ "val_pixel_acc": 0.6,
43
+ "val_precision": 0.6,
44
+ "val_recall": 0.6,
45
+ "val_specificity": 0.6
46
+ }
47
+ }
48
+ ]
versions/v1/checkpoint_best_config.yaml ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ base_batch_size: 16
33
+ num_gpus: 1
34
+ _batch_size_scaled: true
35
+ model:
36
+ name: unet
37
+ in_channels: 3
38
+ out_channels: 1
39
+ features:
40
+ - 8
41
+ - 16
42
+ bilinear: true
43
+ dropout: 0.0
44
+ aux_classifier: false
45
+ num_classes: 3
46
+ gradient_checkpointing: false
47
+ loss:
48
+ mask_loss_type: combined
49
+ bce_weight: 0.5
50
+ dice_weight: 0.5
51
+ focal_weight: 0.5
52
+ focal_gamma: 2.0
53
+ focal_alpha: 0.25
54
+ aux_loss_type: cross_entropy
55
+ aux_weight: 0.2
56
+ training:
57
+ epochs: 1
58
+ learning_rate: 0.001
59
+ weight_decay: 0.0001
60
+ optimizer: adamw
61
+ use_8bit_optimizer: false
62
+ check_8bit_compatibility: true
63
+ fallback_to_16bit: true
64
+ fallback_on_unsupported_8bit: true
65
+ scheduler: cosine
66
+ resume_lr_mode: auto
67
+ resume_lr: null
68
+ warmup_epochs: 1
69
+ min_lr: 1.0e-06
70
+ grad_clip_norm: 1.0
71
+ amp: false
72
+ gradient_accumulation_steps: 1
73
+ data_parallel: false
74
+ use_hard_mining: false
75
+ auto_batch_size: true
76
+ empty_cache_per_epoch: true
77
+ save_best: true
78
+ save_latest: true
79
+ checkpoint_period: 3600
80
+ eval_interval: 1
81
+ early_stopping_patience: 2
82
+ early_stopping_metric: val_iou
83
+ early_stopping_mode: max
84
+ push_to_hub: true
85
+ hub_repo: KhangTruong/Testing-model
86
+ hub_version: v1
87
+ batch_size: 2
88
+ hf_repo: KhangTruong/Testing-model
89
+ num_gpus: 1
90
+ hard_mining:
91
+ enabled: false
92
+ metric: median
93
+ reset_epochs: 5
94
+ bootstrapping:
95
+ enabled: false
96
+ run_bootstrap: false
97
+ epochs: 5
98
+ num_samples: 512
99
+ batch_size: null
100
+ learning_rate: null
101
+ freeze_strategy: channel_stream
102
+ stream_ratio: 0.5
103
+ release_mode: restore
104
+ freeze_modules: []
105
+ initialization: kaiming_normal
106
+ target_score: 0.5
107
+ min_loss_drop: 0.15
108
+ early_stopping: false
109
+ patience: 3
110
+ logging:
111
+ log_interval: 1
112
+ log_memory: true
113
+ measure_network: true
114
+ post_processing:
115
+ enabled: true
116
+ min_area: 64
117
+ fill_holes: true
118
+ max_hole_size: 256
119
+ morphology: open_close
120
+ morph_kernel_size: 3
versions/v1/checkpoint_latest_config.yaml ADDED
@@ -0,0 +1,120 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ base_batch_size: 16
33
+ num_gpus: 1
34
+ _batch_size_scaled: true
35
+ model:
36
+ name: unet
37
+ in_channels: 3
38
+ out_channels: 1
39
+ features:
40
+ - 8
41
+ - 16
42
+ bilinear: true
43
+ dropout: 0.0
44
+ aux_classifier: false
45
+ num_classes: 3
46
+ gradient_checkpointing: false
47
+ loss:
48
+ mask_loss_type: combined
49
+ bce_weight: 0.5
50
+ dice_weight: 0.5
51
+ focal_weight: 0.5
52
+ focal_gamma: 2.0
53
+ focal_alpha: 0.25
54
+ aux_loss_type: cross_entropy
55
+ aux_weight: 0.2
56
+ training:
57
+ epochs: 1
58
+ learning_rate: 0.001
59
+ weight_decay: 0.0001
60
+ optimizer: adamw
61
+ use_8bit_optimizer: false
62
+ check_8bit_compatibility: true
63
+ fallback_to_16bit: true
64
+ fallback_on_unsupported_8bit: true
65
+ scheduler: cosine
66
+ resume_lr_mode: auto
67
+ resume_lr: null
68
+ warmup_epochs: 1
69
+ min_lr: 1.0e-06
70
+ grad_clip_norm: 1.0
71
+ amp: false
72
+ gradient_accumulation_steps: 1
73
+ data_parallel: false
74
+ use_hard_mining: false
75
+ auto_batch_size: true
76
+ empty_cache_per_epoch: true
77
+ save_best: true
78
+ save_latest: true
79
+ checkpoint_period: 3600
80
+ eval_interval: 1
81
+ early_stopping_patience: 2
82
+ early_stopping_metric: val_iou
83
+ early_stopping_mode: max
84
+ push_to_hub: true
85
+ hub_repo: KhangTruong/Testing-model
86
+ hub_version: v1
87
+ batch_size: 2
88
+ hf_repo: KhangTruong/Testing-model
89
+ num_gpus: 1
90
+ hard_mining:
91
+ enabled: false
92
+ metric: median
93
+ reset_epochs: 5
94
+ bootstrapping:
95
+ enabled: false
96
+ run_bootstrap: false
97
+ epochs: 5
98
+ num_samples: 512
99
+ batch_size: null
100
+ learning_rate: null
101
+ freeze_strategy: channel_stream
102
+ stream_ratio: 0.5
103
+ release_mode: restore
104
+ freeze_modules: []
105
+ initialization: kaiming_normal
106
+ target_score: 0.5
107
+ min_loss_drop: 0.15
108
+ early_stopping: false
109
+ patience: 3
110
+ logging:
111
+ log_interval: 1
112
+ log_memory: true
113
+ measure_network: true
114
+ post_processing:
115
+ enabled: true
116
+ min_area: 64
117
+ fill_holes: true
118
+ max_hole_size: 256
119
+ morphology: open_close
120
+ morph_kernel_size: 3
versions/v1/checkpoints/checkpoint_best.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9172bf7d3e6103e1b5a3f72fafb499bd31cef50412504f57501c9bb3bae43a4
3
+ size 217443
versions/v1/checkpoints/checkpoint_latest.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a40f37c060ab46b418b0e7ae2ec1c17db6b65f29ff9b38bfcb64630e174cdbd3
3
+ size 217771
versions/v1/effective_config.yaml ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ project:
2
+ name: test_smoke
3
+ seed: 42
4
+ device: cpu
5
+ output_dir: /tmp/tmp08k8509s/out/RUN/test_smoke
6
+ hub_repo: null
7
+ resume_repo: null
8
+ data:
9
+ dataset_name: KhangTruong/IMD2020
10
+ streaming: false
11
+ image_size:
12
+ - 32
13
+ - 32
14
+ batch_size: 16
15
+ num_workers: 0
16
+ pin_memory: false
17
+ shuffle_buffer_size: 1000
18
+ train_split: train
19
+ val_split: validation
20
+ test_split: test
21
+ eval_split: test
22
+ train_samples_per_epoch: 2
23
+ val_samples_per_epoch: -1
24
+ val_samples: 2
25
+ test_samples: 2
26
+ evaluate_on_test: true
27
+ augmentations:
28
+ horizontal_flip: 0.5
29
+ vertical_flip: 0.0
30
+ random_rotate90: 0.0
31
+ color_jitter: 0.0
32
+ model:
33
+ name: unet
34
+ in_channels: 3
35
+ out_channels: 1
36
+ features:
37
+ - 8
38
+ - 16
39
+ bilinear: true
40
+ dropout: 0.0
41
+ aux_classifier: false
42
+ num_classes: 3
43
+ gradient_checkpointing: false
44
+ loss:
45
+ mask_loss_type: combined
46
+ bce_weight: 0.5
47
+ dice_weight: 0.5
48
+ focal_weight: 0.5
49
+ focal_gamma: 2.0
50
+ focal_alpha: 0.25
51
+ aux_loss_type: cross_entropy
52
+ aux_weight: 0.2
53
+ training:
54
+ epochs: 1
55
+ learning_rate: 0.001
56
+ weight_decay: 0.0001
57
+ optimizer: adamw
58
+ use_8bit_optimizer: false
59
+ check_8bit_compatibility: true
60
+ fallback_to_16bit: true
61
+ fallback_on_unsupported_8bit: true
62
+ scheduler: cosine
63
+ resume_lr_mode: auto
64
+ resume_lr: null
65
+ warmup_epochs: 1
66
+ min_lr: 1.0e-06
67
+ grad_clip_norm: 1.0
68
+ amp: false
69
+ gradient_accumulation_steps: 1
70
+ data_parallel: true
71
+ use_hard_mining: false
72
+ auto_batch_size: true
73
+ empty_cache_per_epoch: true
74
+ save_best: true
75
+ save_latest: true
76
+ checkpoint_period: 3600
77
+ eval_interval: 1
78
+ early_stopping_patience: 2
79
+ early_stopping_metric: val_iou
80
+ early_stopping_mode: max
81
+ push_to_hub: true
82
+ hub_repo: KhangTruong/Testing-model
83
+ hub_version: v1
84
+ batch_size: 2
85
+ hf_repo: KhangTruong/Testing-model
86
+ hard_mining:
87
+ enabled: false
88
+ metric: median
89
+ reset_epochs: 5
90
+ bootstrapping:
91
+ enabled: false
92
+ run_bootstrap: false
93
+ epochs: 5
94
+ num_samples: 512
95
+ batch_size: null
96
+ learning_rate: null
97
+ freeze_strategy: channel_stream
98
+ stream_ratio: 0.5
99
+ release_mode: restore
100
+ freeze_modules: []
101
+ initialization: kaiming_normal
102
+ target_score: 0.5
103
+ min_loss_drop: 0.15
104
+ early_stopping: false
105
+ patience: 3
106
+ logging:
107
+ log_interval: 1
108
+ log_memory: true
109
+ measure_network: true
110
+ post_processing:
111
+ enabled: true
112
+ min_area: 64
113
+ fill_holes: true
114
+ max_hole_size: 256
115
+ morphology: open_close
116
+ morph_kernel_size: 3
versions/v1/manifest.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "versions": [
3
+ "v1"
4
+ ],
5
+ "versions_meta": {
6
+ "v1": {
7
+ "version": "v1",
8
+ "best_score": "0.6",
9
+ "best_epoch": "1",
10
+ "total_epochs": "1",
11
+ "step": "1",
12
+ "updated_at": "2026-10-06T12:25:39.977544+00:00",
13
+ "files": [
14
+ "checkpoint_best.pt",
15
+ "checkpoint_latest.pt"
16
+ ]
17
+ }
18
+ },
19
+ "current_version": "v1"
20
+ }
versions/v1/reports/test_evaluation_report.json ADDED
@@ -0,0 +1,212 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "overall_metrics": {
3
+ "test_bce_loss": 0.6941993832588196,
4
+ "test_dice_loss": 0.6576691269874573,
5
+ "test_focal_loss": 0.09276691824197769,
6
+ "test_mask_loss": 0.7223176956176758,
7
+ "test_total_loss": 0.7223176956176758,
8
+ "test_iou": 0.5,
9
+ "test_dice": 0.5,
10
+ "test_f1": 0.5,
11
+ "test_pixel_f1": 0.5,
12
+ "test_auroc": 0.7407521173590794,
13
+ "test_pixel_auroc": 0.7407521173590794,
14
+ "test_pixel_acc": 0.5,
15
+ "test_precision": 0.5,
16
+ "test_recall": 0.5,
17
+ "test_specificity": 0.5
18
+ },
19
+ "per_label_metrics": {
20
+ "Label 0 (Real / Black Mask)": {
21
+ "iou": 0.0,
22
+ "dice": 0.0,
23
+ "f1": 0.0,
24
+ "pixel_f1": 0.0,
25
+ "auroc": 0.48150423471815884,
26
+ "pixel_auroc": 0.48150423471815884,
27
+ "pixel_acc": 0.0,
28
+ "precision": 0.0,
29
+ "recall": 0.0,
30
+ "specificity": 0.0,
31
+ "samples": 1
32
+ },
33
+ "Label 1 (Synthetic / White Mask)": {
34
+ "iou": 1.0,
35
+ "dice": 1.0,
36
+ "f1": 1.0,
37
+ "pixel_f1": 1.0,
38
+ "auroc": 1.0,
39
+ "pixel_auroc": 1.0,
40
+ "pixel_acc": 1.0,
41
+ "precision": 1.0,
42
+ "recall": 1.0,
43
+ "specificity": 1.0,
44
+ "samples": 1
45
+ },
46
+ "Label 2 (Tampered / Partial Mask)": {
47
+ "iou": 0.0,
48
+ "dice": 0.0,
49
+ "f1": 0.0,
50
+ "pixel_f1": 0.0,
51
+ "auroc": 0.0,
52
+ "pixel_auroc": 0.0,
53
+ "pixel_acc": 0.0,
54
+ "precision": 0.0,
55
+ "recall": 0.0,
56
+ "specificity": 0.0,
57
+ "samples": 0
58
+ }
59
+ },
60
+ "configuration": {
61
+ "project": {
62
+ "name": "test_smoke",
63
+ "seed": 42,
64
+ "device": "cpu",
65
+ "output_dir": "/tmp/tmp08k8509s/out/RUN/test_smoke",
66
+ "hub_repo": null,
67
+ "resume_repo": null
68
+ },
69
+ "data": {
70
+ "dataset_name": "KhangTruong/IMD2020",
71
+ "streaming": false,
72
+ "image_size": [
73
+ 32,
74
+ 32
75
+ ],
76
+ "batch_size": 16,
77
+ "num_workers": 0,
78
+ "pin_memory": false,
79
+ "shuffle_buffer_size": 1000,
80
+ "train_split": "train",
81
+ "val_split": "validation",
82
+ "test_split": "test",
83
+ "eval_split": "test",
84
+ "train_samples_per_epoch": 2,
85
+ "val_samples_per_epoch": -1,
86
+ "val_samples": 2,
87
+ "test_samples": 2,
88
+ "evaluate_on_test": true,
89
+ "augmentations": {
90
+ "horizontal_flip": 0.5,
91
+ "vertical_flip": 0.0,
92
+ "random_rotate90": 0.0,
93
+ "color_jitter": 0.0
94
+ },
95
+ "base_batch_size": 16,
96
+ "num_gpus": 1,
97
+ "_batch_size_scaled": true
98
+ },
99
+ "model": {
100
+ "name": "unet",
101
+ "in_channels": 3,
102
+ "out_channels": 1,
103
+ "features": [
104
+ 8,
105
+ 16
106
+ ],
107
+ "bilinear": true,
108
+ "dropout": 0.0,
109
+ "aux_classifier": false,
110
+ "num_classes": 3,
111
+ "gradient_checkpointing": false
112
+ },
113
+ "loss": {
114
+ "mask_loss_type": "combined",
115
+ "bce_weight": 0.5,
116
+ "dice_weight": 0.5,
117
+ "focal_weight": 0.5,
118
+ "focal_gamma": 2.0,
119
+ "focal_alpha": 0.25,
120
+ "aux_loss_type": "cross_entropy",
121
+ "aux_weight": 0.2
122
+ },
123
+ "training": {
124
+ "epochs": 1,
125
+ "learning_rate": 0.001,
126
+ "weight_decay": 0.0001,
127
+ "optimizer": "adamw",
128
+ "use_8bit_optimizer": false,
129
+ "check_8bit_compatibility": true,
130
+ "fallback_to_16bit": true,
131
+ "fallback_on_unsupported_8bit": true,
132
+ "scheduler": "cosine",
133
+ "resume_lr_mode": "auto",
134
+ "resume_lr": null,
135
+ "warmup_epochs": 1,
136
+ "min_lr": 1e-06,
137
+ "grad_clip_norm": 1.0,
138
+ "amp": false,
139
+ "gradient_accumulation_steps": 1,
140
+ "data_parallel": false,
141
+ "use_hard_mining": false,
142
+ "auto_batch_size": true,
143
+ "empty_cache_per_epoch": true,
144
+ "save_best": true,
145
+ "save_latest": true,
146
+ "checkpoint_period": 3600,
147
+ "eval_interval": 1,
148
+ "early_stopping_patience": 2,
149
+ "early_stopping_metric": "val_iou",
150
+ "early_stopping_mode": "max",
151
+ "push_to_hub": true,
152
+ "hub_repo": "KhangTruong/Testing-model",
153
+ "hub_version": "v1",
154
+ "batch_size": 2,
155
+ "hf_repo": "KhangTruong/Testing-model",
156
+ "num_gpus": 1
157
+ },
158
+ "hard_mining": {
159
+ "enabled": false,
160
+ "metric": "median",
161
+ "reset_epochs": 5
162
+ },
163
+ "bootstrapping": {
164
+ "enabled": false,
165
+ "run_bootstrap": false,
166
+ "epochs": 5,
167
+ "num_samples": 512,
168
+ "batch_size": null,
169
+ "learning_rate": null,
170
+ "freeze_strategy": "channel_stream",
171
+ "stream_ratio": 0.5,
172
+ "release_mode": "restore",
173
+ "freeze_modules": [],
174
+ "initialization": "kaiming_normal",
175
+ "target_score": 0.5,
176
+ "min_loss_drop": 0.15,
177
+ "early_stopping": false,
178
+ "patience": 3
179
+ },
180
+ "logging": {
181
+ "log_interval": 1,
182
+ "log_memory": true,
183
+ "measure_network": true
184
+ },
185
+ "post_processing": {
186
+ "enabled": true,
187
+ "min_area": 64,
188
+ "fill_holes": true,
189
+ "max_hole_size": 256,
190
+ "morphology": "open_close",
191
+ "morph_kernel_size": 3
192
+ }
193
+ },
194
+ "hyperparameters": {
195
+ "Project Name": "test_smoke",
196
+ "Dataset": "KhangTruong/IMD2020",
197
+ "Splits": "Train: 'train' | Val: 'validation' | Test: 'test'",
198
+ "Streaming": false,
199
+ "Image Size": "[32, 32]",
200
+ "Batch Size": 16,
201
+ "Epoch Limit / Steps": "2 samples/epoch",
202
+ "Model": "unet (features=[8, 16], bilinear=True)",
203
+ "Aux Classifier": "False (classes=3)",
204
+ "Mask Loss": "Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5)",
205
+ "Aux Loss": "cross_entropy (wt=0.2)",
206
+ "Post-Processing": "Enabled (min_area=64, fill_holes=True, morph=open_close)",
207
+ "Optimizer": "adamw (lr=0.001, wd=0.0001)",
208
+ "Scheduler": "cosine",
209
+ "Epochs": 1,
210
+ "AMP": false
211
+ }
212
+ }
versions/v1/reports/test_evaluation_report.md ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SID-UNet Evaluation Report
2
+
3
+ ### Run Configuration & Hyperparameters
4
+
5
+ | Hyperparameter / Setting | Configured Value |
6
+ |----------------------------|----------------------------------------------------------|
7
+ | Project Name | test_smoke |
8
+ | Dataset | KhangTruong/IMD2020 |
9
+ | Splits | Train: 'train' | Val: 'validation' | Test: 'test' |
10
+ | Streaming | False |
11
+ | Image Size | [32, 32] |
12
+ | Batch Size | 16 |
13
+ | Epoch Limit / Steps | 2 samples/epoch |
14
+ | Model | unet (features=[8, 16], bilinear=True) |
15
+ | Aux Classifier | False (classes=3) |
16
+ | Mask Loss | Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5) |
17
+ | Aux Loss | cross_entropy (wt=0.2) |
18
+ | Post-Processing | Enabled (min_area=64, fill_holes=True, morph=open_close) |
19
+ | Optimizer | adamw (lr=0.001, wd=0.0001) |
20
+ | Scheduler | cosine |
21
+ | Epochs | 1 |
22
+ | AMP | False |
23
+
24
+ ### Overall Segmentation & Classification Metrics
25
+
26
+ | Metric | Value |
27
+ |------------------|---------|
28
+ | Test Bce Loss | 0.6942 |
29
+ | Test Dice Loss | 0.6577 |
30
+ | Test Focal Loss | 0.0928 |
31
+ | Test Mask Loss | 0.7223 |
32
+ | Test Total Loss | 0.7223 |
33
+ | Test Iou | 0.5 |
34
+ | Test Dice | 0.5 |
35
+ | Test F1 | 0.5 |
36
+ | Test Pixel F1 | 0.5 |
37
+ | Test Auroc | 0.7408 |
38
+ | Test Pixel Auroc | 0.7408 |
39
+ | Test Pixel Acc | 0.5 |
40
+ | Test Precision | 0.5 |
41
+ | Test Recall | 0.5 |
42
+ | Test Specificity | 0.5 |
43
+
44
+ ### Per-Subset Metrics Breakdown
45
+
46
+ | Subset | Mean IoU | Dice / F1 | Pixel Accuracy | Sample Count |
47
+ |-----------------------------------|------------|-------------|------------------|----------------|
48
+ | Label 0 (Real / Black Mask) | 0 | 0 | 0 | 1 |
49
+ | Label 1 (Synthetic / White Mask) | 1 | 1 | 1 | 1 |
50
+ | Label 2 (Tampered / Partial Mask) | 0 | 0 | 0 | 0 |
51
+
versions/v1/reports/training_curves.jpg ADDED

Git LFS Details

  • SHA256: 71a598c9822d6cd64d2fbcf319549f66c7b0b6a6ec810076fe0563f50e1779ae
  • Pointer size: 131 Bytes
  • Size of remote file: 512 kB
versions/v1/reports/training_curves.pdf ADDED
Binary file (38.3 kB). View file
 
versions/v1/reports/training_curves.png ADDED

Git LFS Details

  • SHA256: fc236cd58e55a20de123a51c994b7cdcfd690e1dedb6c41e9abafee038fd6a70
  • Pointer size: 131 Bytes
  • Size of remote file: 362 kB
versions/v1/reports/training_final_report.json ADDED
@@ -0,0 +1,261 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "overall_metrics": {
3
+ "val_bce_loss": 0.6869940757751465,
4
+ "val_dice_loss": 0.590221107006073,
5
+ "val_focal_loss": 0.08142829686403275,
6
+ "val_mask_loss": 0.6793217658996582,
7
+ "val_total_loss": 0.6793217658996582,
8
+ "val_iou": 0.6,
9
+ "val_dice": 0.6,
10
+ "val_f1": 0.6,
11
+ "val_pixel_f1": 0.6,
12
+ "val_auroc": 0.79323464758927,
13
+ "val_pixel_auroc": 0.79323464758927,
14
+ "val_pixel_acc": 0.6,
15
+ "val_precision": 0.6,
16
+ "val_recall": 0.6,
17
+ "val_specificity": 0.6
18
+ },
19
+ "per_label_metrics": {
20
+ "Label 0 (Real / Black Mask)": {
21
+ "iou": 0.0,
22
+ "dice": 0.0,
23
+ "f1": 0.0,
24
+ "pixel_f1": 0.0,
25
+ "auroc": 0.48308661897317506,
26
+ "pixel_auroc": 0.48308661897317506,
27
+ "pixel_acc": 0.0,
28
+ "precision": 0.0,
29
+ "recall": 0.0,
30
+ "specificity": 0.0,
31
+ "samples": 4
32
+ },
33
+ "Label 1 (Synthetic / White Mask)": {
34
+ "iou": 1.0,
35
+ "dice": 1.0,
36
+ "f1": 1.0,
37
+ "pixel_f1": 1.0,
38
+ "auroc": 1.0,
39
+ "pixel_auroc": 1.0,
40
+ "pixel_acc": 1.0,
41
+ "precision": 1.0,
42
+ "recall": 1.0,
43
+ "specificity": 1.0,
44
+ "samples": 3
45
+ },
46
+ "Label 2 (Tampered / Partial Mask)": {
47
+ "iou": 1.0,
48
+ "dice": 1.0,
49
+ "f1": 1.0,
50
+ "pixel_f1": 1.0,
51
+ "auroc": 1.0,
52
+ "pixel_auroc": 1.0,
53
+ "pixel_acc": 1.0,
54
+ "precision": 1.0,
55
+ "recall": 1.0,
56
+ "specificity": 1.0,
57
+ "samples": 3
58
+ }
59
+ },
60
+ "curves_plot_path": "/tmp/tmp08k8509s/out/RUN/test_smoke/reports/training_curves.png",
61
+ "training_history": [
62
+ {
63
+ "epoch": 1,
64
+ "lr": 0.001,
65
+ "train_loss": 0.7411096692085266,
66
+ "train_mask_loss": 0.7411096692085266,
67
+ "train_aux_loss": 0.0,
68
+ "train_iou": 0.17578125,
69
+ "val_loss": 0.6793217658996582,
70
+ "val_mask_loss": 0.6793217658996582,
71
+ "val_aux_loss": 0.0,
72
+ "val_iou": 0.6,
73
+ "val_dice": 0.6,
74
+ "val_f1": 0.6,
75
+ "val_pixel_f1": 0.6,
76
+ "val_auroc": 0.79323464758927,
77
+ "val_pixel_auroc": 0.79323464758927,
78
+ "val_pixel_acc": 0.6,
79
+ "val_precision": 0.6,
80
+ "val_recall": 0.6,
81
+ "val_aux_accuracy": 0.0,
82
+ "train_metrics": {
83
+ "bce_loss": 0.7112534642219543,
84
+ "dice_loss": 0.673093318939209,
85
+ "focal_loss": 0.09787261486053467,
86
+ "mask_loss": 0.7411096692085266,
87
+ "total_loss": 0.7411096692085266,
88
+ "iou": 0.17578125
89
+ },
90
+ "val_metrics": {
91
+ "val_bce_loss": 0.6869940757751465,
92
+ "val_dice_loss": 0.590221107006073,
93
+ "val_focal_loss": 0.08142829686403275,
94
+ "val_mask_loss": 0.6793217658996582,
95
+ "val_total_loss": 0.6793217658996582,
96
+ "val_iou": 0.6,
97
+ "val_dice": 0.6,
98
+ "val_f1": 0.6,
99
+ "val_pixel_f1": 0.6,
100
+ "val_auroc": 0.79323464758927,
101
+ "val_pixel_auroc": 0.79323464758927,
102
+ "val_pixel_acc": 0.6,
103
+ "val_precision": 0.6,
104
+ "val_recall": 0.6,
105
+ "val_specificity": 0.6
106
+ }
107
+ }
108
+ ],
109
+ "configuration": {
110
+ "project": {
111
+ "name": "test_smoke",
112
+ "seed": 42,
113
+ "device": "cpu",
114
+ "output_dir": "/tmp/tmp08k8509s/out/RUN/test_smoke",
115
+ "hub_repo": null,
116
+ "resume_repo": null
117
+ },
118
+ "data": {
119
+ "dataset_name": "KhangTruong/IMD2020",
120
+ "streaming": false,
121
+ "image_size": [
122
+ 32,
123
+ 32
124
+ ],
125
+ "batch_size": 16,
126
+ "num_workers": 0,
127
+ "pin_memory": false,
128
+ "shuffle_buffer_size": 1000,
129
+ "train_split": "train",
130
+ "val_split": "validation",
131
+ "test_split": "test",
132
+ "eval_split": "test",
133
+ "train_samples_per_epoch": 2,
134
+ "val_samples_per_epoch": -1,
135
+ "val_samples": 2,
136
+ "test_samples": 2,
137
+ "evaluate_on_test": true,
138
+ "augmentations": {
139
+ "horizontal_flip": 0.5,
140
+ "vertical_flip": 0.0,
141
+ "random_rotate90": 0.0,
142
+ "color_jitter": 0.0
143
+ },
144
+ "base_batch_size": 16,
145
+ "num_gpus": 1,
146
+ "_batch_size_scaled": true
147
+ },
148
+ "model": {
149
+ "name": "unet",
150
+ "in_channels": 3,
151
+ "out_channels": 1,
152
+ "features": [
153
+ 8,
154
+ 16
155
+ ],
156
+ "bilinear": true,
157
+ "dropout": 0.0,
158
+ "aux_classifier": false,
159
+ "num_classes": 3,
160
+ "gradient_checkpointing": false
161
+ },
162
+ "loss": {
163
+ "mask_loss_type": "combined",
164
+ "bce_weight": 0.5,
165
+ "dice_weight": 0.5,
166
+ "focal_weight": 0.5,
167
+ "focal_gamma": 2.0,
168
+ "focal_alpha": 0.25,
169
+ "aux_loss_type": "cross_entropy",
170
+ "aux_weight": 0.2
171
+ },
172
+ "training": {
173
+ "epochs": 1,
174
+ "learning_rate": 0.001,
175
+ "weight_decay": 0.0001,
176
+ "optimizer": "adamw",
177
+ "use_8bit_optimizer": false,
178
+ "check_8bit_compatibility": true,
179
+ "fallback_to_16bit": true,
180
+ "fallback_on_unsupported_8bit": true,
181
+ "scheduler": "cosine",
182
+ "resume_lr_mode": "auto",
183
+ "resume_lr": null,
184
+ "warmup_epochs": 1,
185
+ "min_lr": 1e-06,
186
+ "grad_clip_norm": 1.0,
187
+ "amp": false,
188
+ "gradient_accumulation_steps": 1,
189
+ "data_parallel": false,
190
+ "use_hard_mining": false,
191
+ "auto_batch_size": true,
192
+ "empty_cache_per_epoch": true,
193
+ "save_best": true,
194
+ "save_latest": true,
195
+ "checkpoint_period": 3600,
196
+ "eval_interval": 1,
197
+ "early_stopping_patience": 2,
198
+ "early_stopping_metric": "val_iou",
199
+ "early_stopping_mode": "max",
200
+ "push_to_hub": true,
201
+ "hub_repo": "KhangTruong/Testing-model",
202
+ "hub_version": "v1",
203
+ "batch_size": 2,
204
+ "hf_repo": "KhangTruong/Testing-model",
205
+ "num_gpus": 1
206
+ },
207
+ "hard_mining": {
208
+ "enabled": false,
209
+ "metric": "median",
210
+ "reset_epochs": 5
211
+ },
212
+ "bootstrapping": {
213
+ "enabled": false,
214
+ "run_bootstrap": false,
215
+ "epochs": 5,
216
+ "num_samples": 512,
217
+ "batch_size": null,
218
+ "learning_rate": null,
219
+ "freeze_strategy": "channel_stream",
220
+ "stream_ratio": 0.5,
221
+ "release_mode": "restore",
222
+ "freeze_modules": [],
223
+ "initialization": "kaiming_normal",
224
+ "target_score": 0.5,
225
+ "min_loss_drop": 0.15,
226
+ "early_stopping": false,
227
+ "patience": 3
228
+ },
229
+ "logging": {
230
+ "log_interval": 1,
231
+ "log_memory": true,
232
+ "measure_network": true
233
+ },
234
+ "post_processing": {
235
+ "enabled": true,
236
+ "min_area": 64,
237
+ "fill_holes": true,
238
+ "max_hole_size": 256,
239
+ "morphology": "open_close",
240
+ "morph_kernel_size": 3
241
+ }
242
+ },
243
+ "hyperparameters": {
244
+ "Project Name": "test_smoke",
245
+ "Dataset": "KhangTruong/IMD2020",
246
+ "Splits": "Train: 'train' | Val: 'validation' | Test: 'test'",
247
+ "Streaming": false,
248
+ "Image Size": "[32, 32]",
249
+ "Batch Size": 16,
250
+ "Epoch Limit / Steps": "2 samples/epoch",
251
+ "Model": "unet (features=[8, 16], bilinear=True)",
252
+ "Aux Classifier": "False (classes=3)",
253
+ "Mask Loss": "Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5)",
254
+ "Aux Loss": "cross_entropy (wt=0.2)",
255
+ "Post-Processing": "Enabled (min_area=64, fill_holes=True, morph=open_close)",
256
+ "Optimizer": "adamw (lr=0.001, wd=0.0001)",
257
+ "Scheduler": "cosine",
258
+ "Epochs": 1,
259
+ "AMP": false
260
+ }
261
+ }
versions/v1/reports/training_final_report.md ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # SID-UNet Evaluation Report
2
+
3
+ ### Run Configuration & Hyperparameters
4
+
5
+ | Hyperparameter / Setting | Configured Value |
6
+ |----------------------------|----------------------------------------------------------|
7
+ | Project Name | test_smoke |
8
+ | Dataset | KhangTruong/IMD2020 |
9
+ | Splits | Train: 'train' | Val: 'validation' | Test: 'test' |
10
+ | Streaming | False |
11
+ | Image Size | [32, 32] |
12
+ | Batch Size | 16 |
13
+ | Epoch Limit / Steps | 2 samples/epoch |
14
+ | Model | unet (features=[8, 16], bilinear=True) |
15
+ | Aux Classifier | False (classes=3) |
16
+ | Mask Loss | Combined (BCE: 0.5, Dice: 0.5, Focal: 0.5) |
17
+ | Aux Loss | cross_entropy (wt=0.2) |
18
+ | Post-Processing | Enabled (min_area=64, fill_holes=True, morph=open_close) |
19
+ | Optimizer | adamw (lr=0.001, wd=0.0001) |
20
+ | Scheduler | cosine |
21
+ | Epochs | 1 |
22
+ | AMP | False |
23
+
24
+ ### Overall Segmentation & Classification Metrics
25
+
26
+ | Metric | Value |
27
+ |-----------------|---------|
28
+ | Val Bce Loss | 0.687 |
29
+ | Val Dice Loss | 0.5902 |
30
+ | Val Focal Loss | 0.0814 |
31
+ | Val Mask Loss | 0.6793 |
32
+ | Val Total Loss | 0.6793 |
33
+ | Val Iou | 0.6 |
34
+ | Val Dice | 0.6 |
35
+ | Val F1 | 0.6 |
36
+ | Val Pixel F1 | 0.6 |
37
+ | Val Auroc | 0.7932 |
38
+ | Val Pixel Auroc | 0.7932 |
39
+ | Val Pixel Acc | 0.6 |
40
+ | Val Precision | 0.6 |
41
+ | Val Recall | 0.6 |
42
+ | Val Specificity | 0.6 |
43
+
44
+ ### Training & Validation Curves
45
+
46
+ ![Training Curves](training_curves.png)
47
+
48
+ ### Epoch-by-Epoch Training & Validation Progression
49
+
50
+ | Epoch | Train Loss | Val Loss | Val IoU | Val F1 | Val AUROC | Val Pixel Acc | Learning Rate |
51
+ |---------|--------------|------------|-----------|----------|-------------|-----------------|-----------------|
52
+ | 1 | 0.7411 | 0.6793 | 0.6 | 0.6 | 0.7932 | 0.6 | 0.001 |
53
+
54
+ ### Per-Subset Metrics Breakdown
55
+
56
+ | Subset | Mean IoU | Dice / F1 | Pixel Accuracy | Sample Count |
57
+ |-----------------------------------|------------|-------------|------------------|----------------|
58
+ | Label 0 (Real / Black Mask) | 0 | 0 | 0 | 4 |
59
+ | Label 1 (Synthetic / White Mask) | 1 | 1 | 1 | 3 |
60
+ | Label 2 (Tampered / Partial Mask) | 1 | 1 | 1 | 3 |
61
+
versions/v1/reports/training_history.csv ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ epoch,lr,train_aux_loss,train_iou,train_loss,train_mask_loss,train_metrics_bce_loss,train_metrics_dice_loss,train_metrics_focal_loss,train_metrics_iou,train_metrics_mask_loss,train_metrics_total_loss,val_auroc,val_aux_accuracy,val_aux_loss,val_dice,val_f1,val_iou,val_loss,val_mask_loss,val_metrics_val_auroc,val_metrics_val_bce_loss,val_metrics_val_dice,val_metrics_val_dice_loss,val_metrics_val_f1,val_metrics_val_focal_loss,val_metrics_val_iou,val_metrics_val_mask_loss,val_metrics_val_pixel_acc,val_metrics_val_pixel_auroc,val_metrics_val_pixel_f1,val_metrics_val_precision,val_metrics_val_recall,val_metrics_val_specificity,val_metrics_val_total_loss,val_pixel_acc,val_pixel_auroc,val_pixel_f1,val_precision,val_recall
2
+ 1,0.001,0.0,0.17578125,0.7411096692085266,0.7411096692085266,0.7112534642219543,0.673093318939209,0.09787261486053467,0.17578125,0.7411096692085266,0.7411096692085266,0.79323464758927,0.0,0.0,0.6,0.6,0.6,0.6793217658996582,0.6793217658996582,0.79323464758927,0.6869940757751465,0.6,0.590221107006073,0.6,0.08142829686403275,0.6,0.6793217658996582,0.6,0.79323464758927,0.6,0.6,0.6,0.6,0.6793217658996582,0.6,0.79323464758927,0.6,0.6,0.6
versions/v1/reports/training_history.json ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "epoch": 1,
4
+ "lr": 0.001,
5
+ "train_loss": 0.7411096692085266,
6
+ "train_mask_loss": 0.7411096692085266,
7
+ "train_aux_loss": 0.0,
8
+ "train_iou": 0.17578125,
9
+ "val_loss": 0.6793217658996582,
10
+ "val_mask_loss": 0.6793217658996582,
11
+ "val_aux_loss": 0.0,
12
+ "val_iou": 0.6,
13
+ "val_dice": 0.6,
14
+ "val_f1": 0.6,
15
+ "val_pixel_f1": 0.6,
16
+ "val_auroc": 0.79323464758927,
17
+ "val_pixel_auroc": 0.79323464758927,
18
+ "val_pixel_acc": 0.6,
19
+ "val_precision": 0.6,
20
+ "val_recall": 0.6,
21
+ "val_aux_accuracy": 0.0,
22
+ "train_metrics": {
23
+ "bce_loss": 0.7112534642219543,
24
+ "dice_loss": 0.673093318939209,
25
+ "focal_loss": 0.09787261486053467,
26
+ "mask_loss": 0.7411096692085266,
27
+ "total_loss": 0.7411096692085266,
28
+ "iou": 0.17578125
29
+ },
30
+ "val_metrics": {
31
+ "val_bce_loss": 0.6869940757751465,
32
+ "val_dice_loss": 0.590221107006073,
33
+ "val_focal_loss": 0.08142829686403275,
34
+ "val_mask_loss": 0.6793217658996582,
35
+ "val_total_loss": 0.6793217658996582,
36
+ "val_iou": 0.6,
37
+ "val_dice": 0.6,
38
+ "val_f1": 0.6,
39
+ "val_pixel_f1": 0.6,
40
+ "val_auroc": 0.79323464758927,
41
+ "val_pixel_auroc": 0.79323464758927,
42
+ "val_pixel_acc": 0.6,
43
+ "val_precision": 0.6,
44
+ "val_recall": 0.6,
45
+ "val_specificity": 0.6
46
+ }
47
+ }
48
+ ]