weiwb commited on
Commit
b27322b
Β·
verified Β·
1 Parent(s): 394a649

Standardize model folders and checkpoint filenames

Browse files
Files changed (31) hide show
  1. README.md +8 -8
  2. {mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/LICENSE +0 -0
  3. {mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/README.md +5 -5
  4. {mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/config.yaml +1 -1
  5. mask2former-cityscapes-olac/model_final.pth β†’ mask2former_cityscapes_olac/mask2former_cityscapes_olac.pth +0 -0
  6. {mask2former-coco-olac β†’ mask2former_coco_olac}/LICENSE +0 -0
  7. {mask2former-coco-olac β†’ mask2former_coco_olac}/README.md +5 -5
  8. {mask2former-coco-olac β†’ mask2former_coco_olac}/config.yaml +1 -1
  9. mask2former-coco-olac/model_final.pth β†’ mask2former_coco_olac/mask2former_coco_olac.pth +0 -0
  10. {pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/LICENSE +0 -0
  11. {pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/README.md +5 -5
  12. {pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/config.yaml +1 -1
  13. pemola-cityscapes-olac/model_final.pth β†’ mask2former_pemola_cityscapes_olac/mask2former_pemola_cityscapes_olac.pth +0 -0
  14. {pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/run_metrics.json +0 -0
  15. {pemola-coco-olac β†’ mask2former_pemola_coco_olac}/LICENSE +0 -0
  16. {pemola-coco-olac β†’ mask2former_pemola_coco_olac}/README.md +5 -5
  17. {pemola-coco-olac β†’ mask2former_pemola_coco_olac}/config.yaml +1 -1
  18. pemola-coco-olac/model_final.pth β†’ mask2former_pemola_coco_olac/mask2former_pemola_coco_olac.pth +0 -0
  19. {pemola-coco-olac β†’ mask2former_pemola_coco_olac}/run_metrics.json +0 -0
  20. {maskdino-coco-olac β†’ maskdino_coco_olac}/LICENSE +0 -0
  21. {maskdino-coco-olac β†’ maskdino_coco_olac}/README.md +4 -4
  22. {maskdino-coco-olac β†’ maskdino_coco_olac}/config.yaml +1 -1
  23. maskdino-coco-olac/model_final.pth β†’ maskdino_coco_olac/maskdino_coco_olac.pth +0 -0
  24. {pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/LICENSE +0 -0
  25. {pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/README.md +4 -4
  26. {pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/config.yaml +1 -1
  27. pemola-maskdino-coco-olac/model_final.pth β†’ maskdino_pemola_coco_olac/maskdino_pemola_coco_olac.pth +0 -0
  28. {pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/LICENSE +0 -0
  29. {pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/README.md +6 -6
  30. {pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/config.yaml +0 -0
  31. pemola-occlusion-swin-l/ep29.pth β†’ swin_l_384_occlusion_coco_olac/swin_l_384_occlusion_coco_olac.pth +0 -0
README.md CHANGED
@@ -12,13 +12,13 @@ All seven model checkpoints, matching configurations, usage notes, and licenses
12
 
13
  | Model | Dataset | Result | Config / Weights |
14
  | :--- | :--- | :---: | :---: |
15
- | [Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former-coco-olac/README.md) | COCO-OLAC | PQ 40.7 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former-coco-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former-coco-olac/model_final.pth) |
16
- | [PEMOLA + Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-coco-olac/README.md) | COCO-OLAC | PQ 41.5 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-coco-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-coco-olac/model_final.pth) |
17
- | [Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former-cityscapes-olac/README.md) | Cityscapes-OLAC | PQ 61.5 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former-cityscapes-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former-cityscapes-olac/model_final.pth) |
18
- | [PEMOLA + Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-cityscapes-olac/README.md) | Cityscapes-OLAC | PQ 62.3 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-cityscapes-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-cityscapes-olac/model_final.pth) |
19
- | [Mask DINO R50](https://huggingface.co/weiwb/PEMOLA/blob/main/maskdino-coco-olac/README.md) | COCO-OLAC | PQ 44.0 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino-coco-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino-coco-olac/model_final.pth) |
20
- | [PEMOLA + Mask DINO R50](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-maskdino-coco-olac/README.md) | COCO-OLAC | PQ 44.8 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-maskdino-coco-olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-maskdino-coco-olac/model_final.pth) |
21
- | [Occlusion classifier: Swin-L 384](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-occlusion-swin-l/README.md) | COCO-OLAC test, black background | Top-1 75.3% | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-occlusion-swin-l/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/pemola-occlusion-swin-l/ep29.pth) |
22
 
23
  ## Download
24
 
@@ -32,7 +32,7 @@ hf download weiwb/PEMOLA --local-dir checkpoints
32
  To download one model, select its folder:
33
 
34
  ```bash
35
- hf download weiwb/PEMOLA --include "pemola-coco-olac/*" --local-dir checkpoints
36
  ```
37
 
38
  For source code, [installation](https://github.com/wenbo-wei/PEMOLA#installation), and [inference](https://github.com/wenbo-wei/PEMOLA#inference--visualization), see the [PEMOLA GitHub repository](https://github.com/wenbo-wei/PEMOLA). Mask DINO models use the dedicated [Mask DINO implementation](https://github.com/wenbo-wei/MaskDINO).
 
12
 
13
  | Model | Dataset | Result | Config / Weights |
14
  | :--- | :--- | :---: | :---: |
15
+ | [Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former_coco_olac/README.md) | COCO-OLAC | PQ 40.7 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_coco_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_coco_olac/mask2former_coco_olac.pth) |
16
+ | [PEMOLA + Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former_pemola_coco_olac/README.md) | COCO-OLAC | PQ 41.5 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_pemola_coco_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_pemola_coco_olac/mask2former_pemola_coco_olac.pth) |
17
+ | [Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former_cityscapes_olac/README.md) | Cityscapes-OLAC | PQ 61.5 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_cityscapes_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_cityscapes_olac/mask2former_cityscapes_olac.pth) |
18
+ | [PEMOLA + Mask2Former R50](https://huggingface.co/weiwb/PEMOLA/blob/main/mask2former_pemola_cityscapes_olac/README.md) | Cityscapes-OLAC | PQ 62.3 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_pemola_cityscapes_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/mask2former_pemola_cityscapes_olac/mask2former_pemola_cityscapes_olac.pth) |
19
+ | [Mask DINO R50](https://huggingface.co/weiwb/PEMOLA/blob/main/maskdino_coco_olac/README.md) | COCO-OLAC | PQ 44.0 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino_coco_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino_coco_olac/maskdino_coco_olac.pth) |
20
+ | [PEMOLA + Mask DINO R50](https://huggingface.co/weiwb/PEMOLA/blob/main/maskdino_pemola_coco_olac/README.md) | COCO-OLAC | PQ 44.8 | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino_pemola_coco_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/maskdino_pemola_coco_olac/maskdino_pemola_coco_olac.pth) |
21
+ | [Occlusion classifier: Swin-L 384](https://huggingface.co/weiwb/PEMOLA/blob/main/swin_l_384_occlusion_coco_olac/README.md) | COCO-OLAC test, black background | Top-1 75.3% | [config](https://huggingface.co/weiwb/PEMOLA/resolve/main/swin_l_384_occlusion_coco_olac/config.yaml) / [weights](https://huggingface.co/weiwb/PEMOLA/resolve/main/swin_l_384_occlusion_coco_olac/swin_l_384_occlusion_coco_olac.pth) |
22
 
23
  ## Download
24
 
 
32
  To download one model, select its folder:
33
 
34
  ```bash
35
+ hf download weiwb/PEMOLA --include "mask2former_pemola_coco_olac/*" --local-dir checkpoints
36
  ```
37
 
38
  For source code, [installation](https://github.com/wenbo-wei/PEMOLA#installation), and [inference](https://github.com/wenbo-wei/PEMOLA#inference--visualization), see the [PEMOLA GitHub repository](https://github.com/wenbo-wei/PEMOLA). Mask DINO models use the dedicated [Mask DINO implementation](https://github.com/wenbo-wei/MaskDINO).
{mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/LICENSE RENAMED
File without changes
{mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/README.md RENAMED
@@ -29,14 +29,14 @@ Run from the PEMOLA repository root after installing its dependencies:
29
  ```bash
30
  python -m pip install huggingface_hub
31
  hf download weiwb/PEMOLA \
32
- --include "mask2former-cityscapes-olac/*" \
33
  --local-dir checkpoints
34
 
35
  python predict.py \
36
- --config checkpoints/mask2former-cityscapes-olac/config.yaml \
37
- --weights checkpoints/mask2former-cityscapes-olac/model_final.pth \
38
  --input /path/to/image.jpg \
39
- --output-dir output/mask2former-cityscapes-olac-predictions \
40
  --no-pemola \
41
  --format png
42
  ```
@@ -45,7 +45,7 @@ The baseline requires only an input image; CAM tensors and occlusion labels are
45
 
46
  ## Files
47
 
48
- - `model_final.pth`: model weights.
49
  - `config.yaml`: matching configuration for the PEMOLA implementation.
50
  - `LICENSE`: MIT license.
51
 
 
29
  ```bash
30
  python -m pip install huggingface_hub
31
  hf download weiwb/PEMOLA \
32
+ --include "mask2former_cityscapes_olac/*" \
33
  --local-dir checkpoints
34
 
35
  python predict.py \
36
+ --config checkpoints/mask2former_cityscapes_olac/config.yaml \
37
+ --weights checkpoints/mask2former_cityscapes_olac/mask2former_cityscapes_olac.pth \
38
  --input /path/to/image.jpg \
39
+ --output-dir output/mask2former_cityscapes_olac-predictions \
40
  --no-pemola \
41
  --format png
42
  ```
 
45
 
46
  ## Files
47
 
48
+ - `mask2former_cityscapes_olac.pth`: model weights.
49
  - `config.yaml`: matching configuration for the PEMOLA implementation.
50
  - `LICENSE`: MIT license.
51
 
{mask2former-cityscapes-olac β†’ mask2former_cityscapes_olac}/config.yaml RENAMED
@@ -383,7 +383,7 @@ MODEL:
383
  QK_SCALE: null
384
  USE_CHECKPOINT: false
385
  WINDOW_SIZE: 7
386
- WEIGHTS: model_final.pth
387
  OUTPUT_DIR: output/mask2former-r50-cityscapes-olac
388
  SEED: 618
389
  SOLVER:
 
383
  QK_SCALE: null
384
  USE_CHECKPOINT: false
385
  WINDOW_SIZE: 7
386
+ WEIGHTS: mask2former_cityscapes_olac.pth
387
  OUTPUT_DIR: output/mask2former-r50-cityscapes-olac
388
  SEED: 618
389
  SOLVER:
mask2former-cityscapes-olac/model_final.pth β†’ mask2former_cityscapes_olac/mask2former_cityscapes_olac.pth RENAMED
File without changes
{mask2former-coco-olac β†’ mask2former_coco_olac}/LICENSE RENAMED
File without changes
{mask2former-coco-olac β†’ mask2former_coco_olac}/README.md RENAMED
@@ -29,14 +29,14 @@ Run from the PEMOLA repository root after installing its dependencies:
29
  ```bash
30
  python -m pip install huggingface_hub
31
  hf download weiwb/PEMOLA \
32
- --include "mask2former-coco-olac/*" \
33
  --local-dir checkpoints
34
 
35
  python predict.py \
36
- --config checkpoints/mask2former-coco-olac/config.yaml \
37
- --weights checkpoints/mask2former-coco-olac/model_final.pth \
38
  --input /path/to/image.jpg \
39
- --output-dir output/mask2former-coco-olac-predictions \
40
  --no-pemola \
41
  --format png
42
  ```
@@ -45,7 +45,7 @@ The baseline requires only an input image; CAM tensors and occlusion labels are
45
 
46
  ## Files
47
 
48
- - `model_final.pth`: model weights.
49
  - `config.yaml`: matching configuration for the PEMOLA implementation.
50
  - `LICENSE`: MIT license.
51
 
 
29
  ```bash
30
  python -m pip install huggingface_hub
31
  hf download weiwb/PEMOLA \
32
+ --include "mask2former_coco_olac/*" \
33
  --local-dir checkpoints
34
 
35
  python predict.py \
36
+ --config checkpoints/mask2former_coco_olac/config.yaml \
37
+ --weights checkpoints/mask2former_coco_olac/mask2former_coco_olac.pth \
38
  --input /path/to/image.jpg \
39
+ --output-dir output/mask2former_coco_olac-predictions \
40
  --no-pemola \
41
  --format png
42
  ```
 
45
 
46
  ## Files
47
 
48
+ - `mask2former_coco_olac.pth`: model weights.
49
  - `config.yaml`: matching configuration for the PEMOLA implementation.
50
  - `LICENSE`: MIT license.
51
 
{mask2former-coco-olac β†’ mask2former_coco_olac}/config.yaml RENAMED
@@ -371,7 +371,7 @@ MODEL:
371
  QK_SCALE: null
372
  USE_CHECKPOINT: false
373
  WINDOW_SIZE: 7
374
- WEIGHTS: model_final.pth
375
  OUTPUT_DIR: output/mask2former-r50-coco-olac
376
  SEED: 618
377
  SOLVER:
 
371
  QK_SCALE: null
372
  USE_CHECKPOINT: false
373
  WINDOW_SIZE: 7
374
+ WEIGHTS: mask2former_coco_olac.pth
375
  OUTPUT_DIR: output/mask2former-r50-coco-olac
376
  SEED: 618
377
  SOLVER:
mask2former-coco-olac/model_final.pth β†’ mask2former_coco_olac/mask2former_coco_olac.pth RENAMED
File without changes
{pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/LICENSE RENAMED
File without changes
{pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/README.md RENAMED
@@ -32,12 +32,12 @@ Install the [PEMOLA environment](https://github.com/wenbo-wei/PEMOLA#installatio
32
 
33
  ```bash
34
  hf download weiwb/PEMOLA \
35
- --include "pemola-cityscapes-olac/*" \
36
  --local-dir checkpoints
37
 
38
  python predict.py \
39
- --config checkpoints/pemola-cityscapes-olac/config.yaml \
40
- --weights checkpoints/pemola-cityscapes-olac/model_final.pth \
41
  --input /path/to/image.png \
42
  --output-dir output/pemola-city-predictions \
43
  --cam-dir /path/to/cam_pt \
@@ -49,11 +49,11 @@ For `image.jpg`, provide `cam_pt/image.pt` and a label JSON entry such as `{"ima
49
 
50
  For Cityscapes, CAM filenames and label keys must include `_leftImg8bit` when the input filename does.
51
 
52
- The [Swin-L classifier](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-occlusion-swin-l/README.md) supplies label-prediction and CAM-generation commands. CAM preparation uses matching original and background-blackened images.
53
 
54
  ## Files
55
 
56
- - `model_final.pth`: model state dictionary under the `model` key.
57
  - `config.yaml`: configuration for this checkpoint.
58
  - `LICENSE`: MIT license.
59
 
 
32
 
33
  ```bash
34
  hf download weiwb/PEMOLA \
35
+ --include "mask2former_pemola_cityscapes_olac/*" \
36
  --local-dir checkpoints
37
 
38
  python predict.py \
39
+ --config checkpoints/mask2former_pemola_cityscapes_olac/config.yaml \
40
+ --weights checkpoints/mask2former_pemola_cityscapes_olac/mask2former_pemola_cityscapes_olac.pth \
41
  --input /path/to/image.png \
42
  --output-dir output/pemola-city-predictions \
43
  --cam-dir /path/to/cam_pt \
 
49
 
50
  For Cityscapes, CAM filenames and label keys must include `_leftImg8bit` when the input filename does.
51
 
52
+ The [Swin-L classifier](https://huggingface.co/weiwb/PEMOLA/blob/main/swin_l_384_occlusion_coco_olac/README.md) supplies label-prediction and CAM-generation commands. CAM preparation uses matching original and background-blackened images.
53
 
54
  ## Files
55
 
56
+ - `mask2former_pemola_cityscapes_olac.pth`: model state dictionary under the `model` key.
57
  - `config.yaml`: configuration for this checkpoint.
58
  - `LICENSE`: MIT license.
59
 
{pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/config.yaml RENAMED
@@ -383,7 +383,7 @@ MODEL:
383
  QK_SCALE: null
384
  USE_CHECKPOINT: false
385
  WINDOW_SIZE: 7
386
- WEIGHTS: model_final.pth
387
  OUTPUT_DIR: ./output
388
  SEED: 618
389
  SOLVER:
 
383
  QK_SCALE: null
384
  USE_CHECKPOINT: false
385
  WINDOW_SIZE: 7
386
+ WEIGHTS: mask2former_pemola_cityscapes_olac.pth
387
  OUTPUT_DIR: ./output
388
  SEED: 618
389
  SOLVER:
pemola-cityscapes-olac/model_final.pth β†’ mask2former_pemola_cityscapes_olac/mask2former_pemola_cityscapes_olac.pth RENAMED
File without changes
{pemola-cityscapes-olac β†’ mask2former_pemola_cityscapes_olac}/run_metrics.json RENAMED
File without changes
{pemola-coco-olac β†’ mask2former_pemola_coco_olac}/LICENSE RENAMED
File without changes
{pemola-coco-olac β†’ mask2former_pemola_coco_olac}/README.md RENAMED
@@ -32,12 +32,12 @@ Install the [PEMOLA environment](https://github.com/wenbo-wei/PEMOLA#installatio
32
 
33
  ```bash
34
  hf download weiwb/PEMOLA \
35
- --include "pemola-coco-olac/*" \
36
  --local-dir checkpoints
37
 
38
  python predict.py \
39
- --config checkpoints/pemola-coco-olac/config.yaml \
40
- --weights checkpoints/pemola-coco-olac/model_final.pth \
41
  --input /path/to/image.jpg \
42
  --output-dir output/pemola-coco-predictions \
43
  --cam-dir /path/to/cam_pt \
@@ -47,11 +47,11 @@ python predict.py \
47
 
48
  For `image.jpg`, provide `cam_pt/image.pt` and a label JSON entry such as `{"image": "mid"}`. CAMs are two-dimensional tensors; labels are `low`, `mid`, or `high`.
49
 
50
- The [Swin-L classifier](https://huggingface.co/weiwb/PEMOLA/blob/main/pemola-occlusion-swin-l/README.md) supplies label-prediction and CAM-generation commands. CAM preparation uses matching original and background-blackened images.
51
 
52
  ## Files
53
 
54
- - `model_final.pth`: model state dictionary under the `model` key.
55
  - `config.yaml`: configuration for this checkpoint.
56
  - `LICENSE`: MIT license.
57
 
 
32
 
33
  ```bash
34
  hf download weiwb/PEMOLA \
35
+ --include "mask2former_pemola_coco_olac/*" \
36
  --local-dir checkpoints
37
 
38
  python predict.py \
39
+ --config checkpoints/mask2former_pemola_coco_olac/config.yaml \
40
+ --weights checkpoints/mask2former_pemola_coco_olac/mask2former_pemola_coco_olac.pth \
41
  --input /path/to/image.jpg \
42
  --output-dir output/pemola-coco-predictions \
43
  --cam-dir /path/to/cam_pt \
 
47
 
48
  For `image.jpg`, provide `cam_pt/image.pt` and a label JSON entry such as `{"image": "mid"}`. CAMs are two-dimensional tensors; labels are `low`, `mid`, or `high`.
49
 
50
+ The [Swin-L classifier](https://huggingface.co/weiwb/PEMOLA/blob/main/swin_l_384_occlusion_coco_olac/README.md) supplies label-prediction and CAM-generation commands. CAM preparation uses matching original and background-blackened images.
51
 
52
  ## Files
53
 
54
+ - `mask2former_pemola_coco_olac.pth`: model state dictionary under the `model` key.
55
  - `config.yaml`: configuration for this checkpoint.
56
  - `LICENSE`: MIT license.
57
 
{pemola-coco-olac β†’ mask2former_pemola_coco_olac}/config.yaml RENAMED
@@ -371,7 +371,7 @@ MODEL:
371
  QK_SCALE: null
372
  USE_CHECKPOINT: false
373
  WINDOW_SIZE: 7
374
- WEIGHTS: model_final.pth
375
  OUTPUT_DIR: ./output
376
  SEED: 618
377
  SOLVER:
 
371
  QK_SCALE: null
372
  USE_CHECKPOINT: false
373
  WINDOW_SIZE: 7
374
+ WEIGHTS: mask2former_pemola_coco_olac.pth
375
  OUTPUT_DIR: ./output
376
  SEED: 618
377
  SOLVER:
pemola-coco-olac/model_final.pth β†’ mask2former_pemola_coco_olac/mask2former_pemola_coco_olac.pth RENAMED
File without changes
{pemola-coco-olac β†’ mask2former_pemola_coco_olac}/run_metrics.json RENAMED
File without changes
{maskdino-coco-olac β†’ maskdino_coco_olac}/LICENSE RENAMED
File without changes
{maskdino-coco-olac β†’ maskdino_coco_olac}/README.md RENAMED
@@ -31,15 +31,15 @@ Use the dedicated [Mask DINO implementation](https://github.com/wenbo-wei/MaskDI
31
 
32
  ```bash
33
  hf download weiwb/PEMOLA \
34
- --include "maskdino-coco-olac/*" \
35
  --local-dir checkpoints
36
 
37
  export DETECTRON2_DATASETS=datasets/data
38
  python train_net.py \
39
- --config-file checkpoints/maskdino-coco-olac/config.yaml \
40
  --num-gpus 1 \
41
  --eval-only \
42
- MODEL.WEIGHTS checkpoints/maskdino-coco-olac/model_final.pth \
43
  DATASETS.TEST '("coco_olac_val",)'
44
  ```
45
 
@@ -47,7 +47,7 @@ The configuration uses the implementation's `MODEL.OAM` options, with `PE_MODULA
47
 
48
  ## Files
49
 
50
- - `model_final.pth`: model state dictionary under the `model` key.
51
  - `config.yaml`: configuration for this checkpoint.
52
  - `LICENSE`: Apache License 2.0, following the Mask DINO source project.
53
 
 
31
 
32
  ```bash
33
  hf download weiwb/PEMOLA \
34
+ --include "maskdino_coco_olac/*" \
35
  --local-dir checkpoints
36
 
37
  export DETECTRON2_DATASETS=datasets/data
38
  python train_net.py \
39
+ --config-file checkpoints/maskdino_coco_olac/config.yaml \
40
  --num-gpus 1 \
41
  --eval-only \
42
+ MODEL.WEIGHTS checkpoints/maskdino_coco_olac/maskdino_coco_olac.pth \
43
  DATASETS.TEST '("coco_olac_val",)'
44
  ```
45
 
 
47
 
48
  ## Files
49
 
50
+ - `maskdino_coco_olac.pth`: model state dictionary under the `model` key.
51
  - `config.yaml`: configuration for this checkpoint.
52
  - `LICENSE`: Apache License 2.0, following the Mask DINO source project.
53
 
{maskdino-coco-olac β†’ maskdino_coco_olac}/config.yaml RENAMED
@@ -379,7 +379,7 @@ MODEL:
379
  QK_SCALE: null
380
  USE_CHECKPOINT: false
381
  WINDOW_SIZE: 7
382
- WEIGHTS: ./model_final.pth
383
  OUTPUT_DIR: ./output/maskdino_r50
384
  SEED: 618
385
  SOLVER:
 
379
  QK_SCALE: null
380
  USE_CHECKPOINT: false
381
  WINDOW_SIZE: 7
382
+ WEIGHTS: ./maskdino_coco_olac.pth
383
  OUTPUT_DIR: ./output/maskdino_r50
384
  SEED: 618
385
  SOLVER:
maskdino-coco-olac/model_final.pth β†’ maskdino_coco_olac/maskdino_coco_olac.pth RENAMED
File without changes
{pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/LICENSE RENAMED
File without changes
{pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/README.md RENAMED
@@ -31,15 +31,15 @@ Use the dedicated [Mask DINO implementation](https://github.com/wenbo-wei/MaskDI
31
 
32
  ```bash
33
  hf download weiwb/PEMOLA \
34
- --include "pemola-maskdino-coco-olac/*" \
35
  --local-dir checkpoints
36
 
37
  export DETECTRON2_DATASETS=datasets/data
38
  python train_net.py \
39
- --config-file checkpoints/pemola-maskdino-coco-olac/config.yaml \
40
  --num-gpus 1 \
41
  --eval-only \
42
- MODEL.WEIGHTS checkpoints/pemola-maskdino-coco-olac/model_final.pth \
43
  DATASETS.TEST '("coco_olac_val",)'
44
  ```
45
 
@@ -49,7 +49,7 @@ The configuration uses the implementation's `MODEL.OAM` options, with `PE_MODULA
49
 
50
  ## Files
51
 
52
- - `model_final.pth`: model state dictionary under the `model` key.
53
  - `config.yaml`: configuration for this checkpoint.
54
  - `LICENSE`: Apache License 2.0, following the Mask DINO source project.
55
 
 
31
 
32
  ```bash
33
  hf download weiwb/PEMOLA \
34
+ --include "maskdino_pemola_coco_olac/*" \
35
  --local-dir checkpoints
36
 
37
  export DETECTRON2_DATASETS=datasets/data
38
  python train_net.py \
39
+ --config-file checkpoints/maskdino_pemola_coco_olac/config.yaml \
40
  --num-gpus 1 \
41
  --eval-only \
42
+ MODEL.WEIGHTS checkpoints/maskdino_pemola_coco_olac/maskdino_pemola_coco_olac.pth \
43
  DATASETS.TEST '("coco_olac_val",)'
44
  ```
45
 
 
49
 
50
  ## Files
51
 
52
+ - `maskdino_pemola_coco_olac.pth`: model state dictionary under the `model` key.
53
  - `config.yaml`: configuration for this checkpoint.
54
  - `LICENSE`: Apache License 2.0, following the Mask DINO source project.
55
 
{pemola-maskdino-coco-olac β†’ maskdino_pemola_coco_olac}/config.yaml RENAMED
@@ -379,7 +379,7 @@ MODEL:
379
  QK_SCALE: null
380
  USE_CHECKPOINT: false
381
  WINDOW_SIZE: 7
382
- WEIGHTS: ./model_final.pth
383
  OUTPUT_DIR: ./output/pemola_maskdino_r50
384
  SEED: 618
385
  SOLVER:
 
379
  QK_SCALE: null
380
  USE_CHECKPOINT: false
381
  WINDOW_SIZE: 7
382
+ WEIGHTS: ./maskdino_pemola_coco_olac.pth
383
  OUTPUT_DIR: ./output/pemola_maskdino_r50
384
  SEED: 618
385
  SOLVER:
pemola-maskdino-coco-olac/model_final.pth β†’ maskdino_pemola_coco_olac/maskdino_pemola_coco_olac.pth RENAMED
File without changes
{pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/LICENSE RENAMED
File without changes
{pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/README.md RENAMED
@@ -33,7 +33,7 @@ Install the [PEMOLA environment](https://github.com/wenbo-wei/PEMOLA#installatio
33
 
34
  ```bash
35
  hf download weiwb/PEMOLA \
36
- --include "pemola-occlusion-swin-l/*" \
37
  --local-dir checkpoints
38
 
39
  mkdir -p output/occlusion-inference
@@ -41,8 +41,8 @@ printf '{}\n' > output/occlusion-inference/empty_labels.json
41
 
42
  torchrun --standalone --nnodes=1 --nproc_per_node=1 \
43
  occ_cls_predict.py \
44
- --cfg checkpoints/pemola-occlusion-swin-l/config.yaml \
45
- --resume checkpoints/pemola-occlusion-swin-l/ep29.pth \
46
  --data_path /path/to/images_blackbg \
47
  --occlusion_ann output/occlusion-inference/empty_labels.json \
48
  --batch_size 1 \
@@ -61,8 +61,8 @@ Provide matching filenames in `/path/to/images` and `/path/to/images_blackbg`, p
61
  ```bash
62
  torchrun --standalone --nnodes=1 --nproc_per_node=1 \
63
  occ_cls_draw_cam.py \
64
- --cfg checkpoints/pemola-occlusion-swin-l/config.yaml \
65
- --resume checkpoints/pemola-occlusion-swin-l/ep29.pth \
66
  --dataset coco \
67
  --data_path /path/to/images \
68
  --occlusion_ann /path/to/occlusion_labels.json \
@@ -75,7 +75,7 @@ CAM tensors are written to `output/occlusion-inference/swin_large_patch4_window1
75
 
76
  ## Files
77
 
78
- - `ep29.pth`: model state dictionary under the `model` key.
79
  - `config.yaml`: classifier configuration.
80
  - `LICENSE`: MIT license.
81
 
 
33
 
34
  ```bash
35
  hf download weiwb/PEMOLA \
36
+ --include "swin_l_384_occlusion_coco_olac/*" \
37
  --local-dir checkpoints
38
 
39
  mkdir -p output/occlusion-inference
 
41
 
42
  torchrun --standalone --nnodes=1 --nproc_per_node=1 \
43
  occ_cls_predict.py \
44
+ --cfg checkpoints/swin_l_384_occlusion_coco_olac/config.yaml \
45
+ --resume checkpoints/swin_l_384_occlusion_coco_olac/swin_l_384_occlusion_coco_olac.pth \
46
  --data_path /path/to/images_blackbg \
47
  --occlusion_ann output/occlusion-inference/empty_labels.json \
48
  --batch_size 1 \
 
61
  ```bash
62
  torchrun --standalone --nnodes=1 --nproc_per_node=1 \
63
  occ_cls_draw_cam.py \
64
+ --cfg checkpoints/swin_l_384_occlusion_coco_olac/config.yaml \
65
+ --resume checkpoints/swin_l_384_occlusion_coco_olac/swin_l_384_occlusion_coco_olac.pth \
66
  --dataset coco \
67
  --data_path /path/to/images \
68
  --occlusion_ann /path/to/occlusion_labels.json \
 
75
 
76
  ## Files
77
 
78
+ - `swin_l_384_occlusion_coco_olac.pth`: model state dictionary under the `model` key.
79
  - `config.yaml`: classifier configuration.
80
  - `LICENSE`: MIT license.
81
 
{pemola-occlusion-swin-l β†’ swin_l_384_occlusion_coco_olac}/config.yaml RENAMED
File without changes
pemola-occlusion-swin-l/ep29.pth β†’ swin_l_384_occlusion_coco_olac/swin_l_384_occlusion_coco_olac.pth RENAMED
File without changes