current_works_v2: baseline_server

current_works_v2/baseline_server/2050/none_accuracy_analysis.txt

@@ -0,0 +1,3 @@
+acc & L/R: 0.98 & F/B: 0.99 & Bi/Sm: 0.96 & Ta/Sh: 0.96 & Stand: 1.00 & Close: 0.74 & Symm: 0.53. Total: &0.96
+means of mean: 0.88
+

current_works_v2/baseline_server/args.json

@@ -0,0 +1,31 @@
+{
+  "batchSize": 60,
+  "nepoch": 10000,
+  "outf": "checkpoint",
+  "model": "",
+  "dataset": "/home/lehoangan/LHAN/SteerScene/echoscene_sizefix/FRONT",
+  "logf": "logs",
+  "exp": "/home/lehoangan/LHAN/SteerScene/echoscene_sizefix/exp_baseline",
+  "room_type": "all",
+  "residual": true,
+  "pooling": "avg",
+  "large": false,
+  "use_scene_rels": true,
+  "separated": false,
+  "with_SDF": true,
+  "with_CLIP": true,
+  "shuffle_objs": true,
+  "use_canonical": true,
+  "with_angles": true,
+  "bin_angle": false,
+  "num_box_params": 6,
+  "workers": 10,
+  "with_changes": true,
+  "loadmodel": false,
+  "loadepoch": 90,
+  "replace_latent": true,
+  "network_type": "echoscene",
+  "diff_yaml": "/home/lehoangan/LHAN/SteerScene/echoscene_sizefix/code/config/baseline_noguidance.yaml",
+  "vis_num": 2,
+  "manualSeed": 8886
+}

current_works_v2/baseline_server/baseline_noguidance.yaml

@@ -0,0 +1,130 @@
+# EchoScene baseline: work41_sizefix.yaml with inference guidance disabled.
+hyper:
+    batch_size: 16
+    gpu_ids: 0
+    logs_dir: ../full_mp/diff_crossattn
+    results_dir: ../full_mp/diff_crossattn
+    name: ./
+    isTrain: True
+    device: 'cuda'
+    distributed: 0
+    lr_init: 1e-4
+    lr_step: [35000, 70000, 140000]
+    lr_evo: [5e-5, 1e-5, 5e-6]
+dataset:
+    res: 64
+    trunc_thres: 0.2
+    ratio: 1
+
+layout_branch:
+    model: diffusion_scene_layout_ddpm
+    angle_dim: 2
+    denoiser: unet1d
+    relation_condition: true
+    denoiser_kwargs:
+        dims: 1 # 1D
+        in_channels: 8 # size(3)+loc(3)+sincos(2)
+        out_channels: 8 # same
+        model_channels: 512
+        channel_mult: [ 1,1,1,1]
+        num_res_blocks: 2
+        attention_resolutions: [ 4, 2 ]
+        num_heads: 8
+        # cond_model params
+        use_spatial_transformer: true
+        transformer_depth: 1
+        conditioning_key: 'crossattn'
+        concat_dim: 1280
+        crossattn_dim: 1280
+        use_checkpoint: true
+        enable_t_emb: true
+
+    diffusion_kwargs:
+      schedule_type: 'linear'
+      beta_start: 0.0001
+      beta_end: 0.02
+      time_num: 1000
+      model_mean_type: 'eps'
+      model_var_type: 'fixedsmall'
+      loss_separate: true
+      loss_iou: false
+      iou_type: obb
+      train_stats_file: null
+    inference_guidance:
+      enabled: false
+      guide_sizes: false  # size-freeze fix: guidance moves/rotates boxes only
+      interval: 1
+      start_ratio: 0.90
+      grad_clip: 10.1
+      clip_to_bounds: true
+      constraints:
+        collision:
+          enabled: true
+          strength: 20.0
+          weight: 10.0
+          iou_threshold: 0.02
+          penetration_threshold: 0.0
+          iou_weight: 0.5
+          penetration_weight: 2.0
+          metric: aabb
+          ignore_labels: ["scene", "floor"]
+        room_outer:
+          weight: 10.0
+        walkable:
+          enabled: false
+          weight: 1.0 # Global scale factor for walkable guidance
+          
+          # Modular sub-components configured to match real_current_best behavior:
+          components:
+            center_penalty:
+              enabled: true
+              weight: 1
+              sigma: 0.5
+              
+            pathfinding:
+              enabled: false
+              weight: 0.1
+              robot_width_real: 0.5
+              robot_hight_real: 1.5
+              
+            edge_gaussian:
+              enabled: true
+              weight: 1
+              robot_width_real: 0.35
+              robot_hight_real: 1.5
+              sigma_scale: 0.5
+              heatmap_weight: 0.8    # Component 1 (Floor Edge Heatmap)
+              repulsion_weight: 0.2   # Component 2 (Pairwise Furniture Repulsion)
+        relational:
+          enabled: true
+          weight: 10.0
+          start_ratio: 0.5      # Custom start ratio for relational guidance (e.g., 0.5, 0.7, 0.95; defaults to global start_ratio if omitted)
+          interval: 1           # Custom interval for relational guidance (e.g., 1=every step, 5=every 5th step; defaults to global interval if omitted)
+          margin: 0.05
+          close_threshold: 0.45
+          stand_threshold: 0.04
+
+shape_branch:
+    model: sdfusion-txt2shape_mp
+    sampling: greedy
+    ckpt: null
+    df_cfg: ../config/sdfusion-txt2shape_mp.yaml
+    ddim_steps: 100
+    ddim_eta: 0.0
+    uc_scale: 3.0
+    vq_model: vqvae
+    vq_cfg: ../config/vqvae_snet.yaml
+    vq_dset: None
+    vq_cat: None
+    vq_ckpt: ./checkpoint/vqvae_threedfront_best.pth
+misc:
+    debug: 0
+    seed: 111
+    backend: gloo
+    local_rank: 0
+
+training:
+    lr: 1e-5
+    lr_policy: lambda
+    lr_decay_iters: 50
+    lambda_L1: 10.0

current_works_v2/baseline_server/eval/pp_col.txt

@@ -0,0 +1,17 @@
+===========================================================================
+          PHYSICAL SCENE COLLISION EVALUATION (Col Scene & Col Obj)          
+===========================================================================
+ Target File                               : /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/physcene_collision_resolved.json
+ Overall ColObj   (Object Collision Rate) : 0.0300 (76/2535 objects)
+ Overall ColScene (Scene Collision Rate)  : 0.0838 (31/370 scenes)
+---------------------------------------------------------------------------
+ Category           | ColObj   | ColScene | Collided Scenes / Total
+---------------------------------------------------------------------------
+ bedroom            | 0.0591   | 0.1220   | 5/41
+ diningroom         | 0.0073   | 0.0290   | 2/69
+ library            | 0.0543   | 0.1071   | 6/56
+ livingdiningroom   | 0.0151   | 0.0323   | 1/31
+ livingroom         | 0.0120   | 0.0385   | 2/52
+ masterbedroom      | 0.0497   | 0.1159   | 8/69
+ secondbedroom      | 0.0600   | 0.1346   | 7/52
+===========================================================================

current_works_v2/baseline_server/eval/pp_ob.txt

@@ -0,0 +1,5 @@
+Total number of scenes: 370 (processed 370)
+Total OB objects: 102
+Total furniture objects: 2527
+Average #OB per scene: 0.2757
+OB Object Rate (tot_ob / tot_obj): 0.0404

current_works_v2/baseline_server/eval/pp_rel.txt

@@ -0,0 +1,17 @@
+Wrote report to: /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/relational_accuracy_report.txt
+===========================================================================
+          RELATIONAL ACCURACY EVALUATION REPORT          
+===========================================================================
+ Source File                            : /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/physcene_collision_resolved.json
+ Evaluated Scenes                       : 370 / 370
+ Micro-Average Accuracy (Total Acc)     : 96.26%
+ Macro-Average Accuracy (Means of Means): 86.59%
+---------------------------------------------------------------------------
+ Left / Right (L/R)       : 98.66%  (Left: 98.42%, Right: 98.91%)
+ Front / Behind (F/B)     : 99.88%  (Front: 99.93%, Behind: 99.83%)
+ Bigger / Smaller (Bi/Sm) : 96.25%  (Bigger: 96.42%, Smaller: 96.09%)
+ Taller / Shorter (Ta/Sh) : 96.22%  (Taller: 96.34%, Shorter: 96.10%)
+ Standing On              : 99.71%
+ Close By                 : 70.60%
+ Symmetrical To           : 44.83%
+===========================================================================

current_works_v2/baseline_server/eval/raw_col.txt

@@ -0,0 +1,17 @@
+===========================================================================
+          PHYSICAL SCENE COLLISION EVALUATION (Col Scene & Col Obj)          
+===========================================================================
+ Target File                               : /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/physcene_collision_input.json
+ Overall ColObj   (Object Collision Rate) : 0.4856 (1231/2535 objects)
+ Overall ColScene (Scene Collision Rate)  : 0.8216 (304/370 scenes)
+---------------------------------------------------------------------------
+ Category           | ColObj   | ColScene | Collided Scenes / Total
+---------------------------------------------------------------------------
+ bedroom            | 0.4089   | 0.7073   | 29/41
+ diningroom         | 0.6245   | 0.9420   | 65/69
+ library            | 0.3992   | 0.6250   | 35/56
+ livingdiningroom   | 0.4497   | 0.9677   | 30/31
+ livingroom         | 0.4157   | 0.7692   | 40/52
+ masterbedroom      | 0.4869   | 0.8841   | 61/69
+ secondbedroom      | 0.5280   | 0.8462   | 44/52
+===========================================================================

current_works_v2/baseline_server/eval/raw_ob.txt

@@ -0,0 +1,5 @@
+Total number of scenes: 370 (processed 370)
+Total OB objects: 443
+Total furniture objects: 2527
+Average #OB per scene: 1.1973
+OB Object Rate (tot_ob / tot_obj): 0.1753

current_works_v2/baseline_server/eval/raw_rel.txt

@@ -0,0 +1,17 @@
+Wrote report to: /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/relational_accuracy_report.txt
+===========================================================================
+          RELATIONAL ACCURACY EVALUATION REPORT          
+===========================================================================
+ Source File                            : /Volumes/SPINNINGHDD/current_works_v2/baseline_server/2050/physcene_collision_input.json
+ Evaluated Scenes                       : 370 / 370
+ Micro-Average Accuracy (Total Acc)     : 96.26%
+ Macro-Average Accuracy (Means of Means): 88.07%
+---------------------------------------------------------------------------
+ Left / Right (L/R)       : 98.35%  (Left: 98.23%, Right: 98.47%)
+ Front / Behind (F/B)     : 99.15%  (Front: 99.19%, Behind: 99.10%)
+ Bigger / Smaller (Bi/Sm) : 96.25%  (Bigger: 96.42%, Smaller: 96.09%)
+ Taller / Shorter (Ta/Sh) : 96.22%  (Taller: 96.34%, Shorter: 96.10%)
+ Standing On              : 99.71%
+ Close By                 : 74.20%
+ Symmetrical To           : 52.59%
+===========================================================================

current_works_v2/baseline_server/vis_2050.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c998159347727486d042b2ae52970a67e0ca904a229d952d95efdc1d88096f3
+size 1462848426