diff --git a/.gitattributes b/.gitattributes index c5edaa3a09d70a96866d58aecb8cb789b66e361b..38eec5eccb890c6bcceb7e687124162369acd7b2 100644 --- a/.gitattributes +++ b/.gitattributes @@ -754,3 +754,7 @@ sessions/session_20260921_225630_786148/preprocessing/trellis-20260925-rejected/ sessions/session_20260921_225630_786148/preprocessing/trellis-20260925-rejected/native/hybrid/loader_smoke/session_20260921_225630_786148/points3D.ply filter=lfs diff=lfs merge=lfs -text sessions/session_20260921_225630_786148/preprocessing/trellis-20260925-rejected/native/shape/object.glb filter=lfs diff=lfs merge=lfs -text sessions/session_20260921_225630_786148/preprocessing/trellis-20260925-rejected/native/shape/object.ply filter=lfs diff=lfs merge=lfs -text +sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1/input_case/session_20260921_225744_574043/observation.ply filter=lfs diff=lfs merge=lfs -text +sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1/input_case/session_20260921_225744_574043/points3D.ply filter=lfs diff=lfs merge=lfs -text +sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1/model/point_cloud/iteration_10000/point_cloud.ply filter=lfs diff=lfs merge=lfs -text +sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1/model/point_cloud/iteration_7000/point_cloud.ply filter=lfs diff=lfs merge=lfs -text diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/APPEARANCE_REVIEW_c6a_v1.md b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/APPEARANCE_REVIEW_c6a_v1.md new file mode 100644 index 0000000000000000000000000000000000000000..e3dc390b291c1c2a60913e0ccfb71c134f2b6868 --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/APPEARANCE_REVIEW_c6a_v1.md @@ -0,0 +1,27 @@ +# 225744: static appearance review + +**Status: reference training and agent review complete; owner appearance accepted.** + +Owner accepted this exact saved artifact at 2026-10-28T17:46:32.866732+00:00. [Acceptance record](../owner_appearance_acceptance_c6a_v1.json). The disclosed limitations remain. + +This appearance uses the accepted 335–775 window, masks and orange prior. It is trained on the four initial reference images with the unchanged recipe: 10,000 loop iterations and 9,999 Gaussian optimizer updates. The prior packet retains each camera’s own capture mapping. No physical fitting has run. + +[Rotation video](appearance_train_v1/review/novel_view.mp4) · [loss curves](appearance_train_v1/review/loss_curves.png) + +- [Camera 12540311: observed/rendered RGB, mask/alpha and depth](appearance_train_v1/review/iteration_10000/view_0_sheet.png) +- [Camera 12933256: observed/rendered RGB, mask/alpha and depth](appearance_train_v1/review/iteration_10000/view_1_sheet.png) +- [Camera 17736741: observed/rendered RGB, mask/alpha and depth](appearance_train_v1/review/iteration_10000/view_2_sheet.png) +- [Camera 23007103: observed/rendered RGB, mask/alpha and depth](appearance_train_v1/review/iteration_10000/view_3_sheet.png) + +The visible orange belt and purple tape align across all four views. **The hand-hidden grasp arc remains incomplete**, with an open gap and fuzzy endpoint/fringe, particularly visible in cameras 12540311 and 17736741. Small endpoint protrusions remain in 12933256 and 23007103. The gap persists in the sampled rotation views; no overall collapse or broad detached cloud was observed. These limitations are preserved in the saved result. The raw residual display compares against the full scene and is not the masked training objective. + +- Initial hybrid seeds: 69,946. Final saved Gaussians: 76,863; reference export after strict opacity >0.1: 67,614. +- One isotropic scale column; final spherical-harmonic degree 3. +- Mean same-view source total: 0.005847150 initially → 0.000194653 at iteration 10000. This averages the source full-image losses over the four training views; final review uses the reference opacity filter. It is not held-out evidence or a physical-accuracy measurement. +- All arrays in 12 saved view files at iterations 0/7000/10000 and both input/final PLY payloads are finite. All three scalar histories cover every step 1–10000 with finite values. +- Training and review returned zero; elapsed wall times 446.69 s and 322.53 s include shared resource-lane waiting. +- [Summary](appearance_train_v1/review/summary.json) · [agent findings and exact coverage](appearance_train_v1/review/agent_findings.json) · [run flags](appearance_train_v1/run_record.json). + +The four final sheets, loss curves and orbit frames 0,25,50,75,100,125,150 and 175 of 200 were visually inspected. The curves settle after transients near 3000/6000/9000, consistent with the unchanged source opacity-reset schedule. Sampled inspection is not certification of every orbit frame. + +Appearance acceptance is separate from prior acceptance and does not authorize physical fitting. diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1.log b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1.log new file mode 100644 index 0000000000000000000000000000000000000000..e6004420aca5ad4c049df1c4c19d060eb1f4485c --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1.log @@ -0,0 +1,57 @@ +Found metadata.json file, assuming customized QQTT dataset! [28/10 13:29:50] +Init points from pcd... [28/10 13:29:51] +Init points from mesh... [28/10 13:29:51] +Loading Training Cameras [28/10 13:29:51] +Loading Test Cameras [28/10 13:29:51] +Number of points at initialisation : 69946 [28/10 13:29:51] +Sampling 100000 points from mesh [28/10 13:29:51] +{ + "cameras": [ + { + "id": 0, + "width": 1920, + "height": 1080, + "object_px": 8480, + "occlusion_px": 145591, + "valid_depth_px": 1980482, + "boundary_px": 3265, + "projection_max_error_px": 2.2288144272505374e-05 + }, + { + "id": 1, + "width": 1920, + "height": 1080, + "object_px": 8696, + "occlusion_px": 137749, + "valid_depth_px": 2025824, + "boundary_px": 3410, + "projection_max_error_px": 5.990189657605037e-07 + }, + { + "id": 2, + "width": 1920, + "height": 1080, + "object_px": 8751, + "occlusion_px": 208201, + "valid_depth_px": 1982256, + "boundary_px": 3396, + "projection_max_error_px": 4.9956752213924245e-06 + }, + { + "id": 3, + "width": 1920, + "height": 1080, + "object_px": 5933, + "occlusion_px": 147479, + "valid_depth_px": 1906085, + "boundary_px": 2774, + "projection_max_error_px": 3.0847624918624304e-06 + } + ], + "hybrid": { + "vertex_color_branch": true, + "mesh": 51840, + "observed": 18106, + "total": 69946 + } +} [28/10 13:29:51] diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_argv.json b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_argv.json new file mode 100644 index 0000000000000000000000000000000000000000..cded885b06fa20ad6610f9f18fad03d4a199a8bd --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_argv.json @@ -0,0 +1,18 @@ +{ + "argv": [ + "/mnt/data0/magna/belt_perception_wt_c6a_lab_input_preparation_20261027/scripts/real_rgbd_gs_train_env/.venv/bin/python", + "scripts/real_rgbd_appearance_train.py", + "check", + "--prior", + "/mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/prior_orange_v1", + "--accepted-prior-manifest-sha256", + "1700148fe9bddb939b16cd2d3a22dfab78165af590b77445dfa63796f1fc01d9", + "--output", + "/mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/appearance_check_v1" + ], + "cwd": "/mnt/data0/magna/belt_perception_wt_c6a_lab_input_preparation_20261027", + "environment": { + "PYTHONDONTWRITEBYTECODE": "1", + "MPLBACKEND": "Agg" + } +} diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_result.json b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_result.json new file mode 100644 index 0000000000000000000000000000000000000000..e502ebc4102f83072d5221dd02879601086e9b81 --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_check_v1_result.json @@ -0,0 +1,4 @@ +{ + "returncode": 0, + "elapsed_s": 301.16960447700694 +} diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1.log b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1.log new file mode 100644 index 0000000000000000000000000000000000000000..9d6dfcfe3270432f5febca2719a457abe42abe96 --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1.log @@ -0,0 +1,32 @@ +/mnt/data0/magna/belt_perception_wt_c6a_lab_input_preparation_20261027/scripts/real_rgbd_gs_train_env/.venv/lib/python3.10/site-packages/torch/utils/cpp_extension.py:1965: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. +If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST']. + warnings.warn( +Found metadata.json file, assuming customized QQTT dataset! [28/10 13:42:29] +Init points from pcd... [28/10 13:42:30] +Init points from mesh... [28/10 13:42:30] +Loading Training Cameras [28/10 13:42:30] +Loading Test Cameras [28/10 13:42:30] +Number of points at initialisation : 69946 [28/10 13:42:30] +Sampling 100000 points from mesh [28/10 13:42:30] +Loading trained model at iteration 7000 [28/10 13:42:32] [28/10 13:42:32] +Found metadata.json file, assuming customized QQTT dataset! [28/10 13:42:32] [28/10 13:42:32] +Init points from pcd... [28/10 13:42:32] [28/10 13:42:32] +Init points from mesh... [28/10 13:42:32] [28/10 13:42:32] +Loading Training Cameras [28/10 13:42:32] [28/10 13:42:32] +Loading Test Cameras [28/10 13:42:32] [28/10 13:42:32] +Sampling 100000 points from mesh [28/10 13:42:32] [28/10 13:42:32] +Loading trained model at iteration 10000 [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Found metadata.json file, assuming customized QQTT dataset! [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Init points from pcd... [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Init points from mesh... [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Loading Training Cameras [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Loading Test Cameras [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +Sampling 100000 points from mesh [28/10 13:42:34] [28/10 13:42:34] [28/10 13:42:34] +{ + "snapshots": [ + "0", + "7000", + "10000" + ], + "review": "/mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/appearance_train_v1/review" +} [28/10 13:42:40] [28/10 13:42:40] [28/10 13:42:40] diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_argv.json b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_argv.json new file mode 100644 index 0000000000000000000000000000000000000000..45abd2d6f70839dbf188bea77ae6a12a575d072a --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_argv.json @@ -0,0 +1,13 @@ +{ + "argv": [ + "/mnt/data0/magna/belt_perception_wt_c6a_lab_input_preparation_20261027/scripts/real_rgbd_gs_train_env/.venv/bin/python", + "scripts/real_rgbd_appearance_review.py", + "--run", + "/mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/appearance_train_v1" + ], + "cwd": "/mnt/data0/magna/belt_perception_wt_c6a_lab_input_preparation_20261027", + "environment": { + "PYTHONDONTWRITEBYTECODE": "1", + "MPLBACKEND": "Agg" + } +} diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_result.json b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_result.json new file mode 100644 index 0000000000000000000000000000000000000000..afc6cb893f789e5516306e44cada31ea64bc4b20 --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_review_v1_result.json @@ -0,0 +1,4 @@ +{ + "returncode": 0, + "elapsed_s": 322.53122205000545 +} diff --git a/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1.log b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1.log new file mode 100644 index 0000000000000000000000000000000000000000..fe776516d8e813647e47480791055ff83cddfbf4 --- /dev/null +++ b/sessions/session_20260921_225744_574043/preprocessing/c6a-input-preparation-v1/native/appearance_train_v1.log @@ -0,0 +1,19 @@ +Optimizing /mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/appearance_train_v1/model +Output folder: /mnt/data2/magna/belt_perception_shared/phystwin_model_corpus_expansion/lab/session_20260921_225744_574043/window_anchor0335-0775/accepted_input_preparation_c6a_v1/appearance_train_v1/model [28/10 13:35:14] +Found metadata.json file, assuming customized QQTT dataset! [28/10 13:35:14] +Init points from pcd... [28/10 13:35:15] +Init points from mesh... [28/10 13:35:15] +Loading Training Cameras [28/10 13:35:15] +Loading Test Cameras [28/10 13:35:15] +Number of points at initialisation : 69946 [28/10 13:35:15] +Sampling 100000 points from mesh [28/10 13:35:15] + Training progress: 0%| | 0/10000 [00:00