lab47d
session
work
/ihub/homedirs/svs_ald/sudhir/real2sim/captures/session_20260902_184520work
/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47dVideo
Stages
10_pose — Pose
Camera poses from unposed images (VGGT / COLMAP / GLOMAP).
How to read it. `registered` should equal `requested`. `n_models` > 1 means the capture split into disconnected pieces -- not enough overlap. A near-zero baseline means the camera barely moved.
| backend | colmap |
|---|---|
| views | 47 |
| baseline_recon_units | 9.436 |
| points | 23,825 |
| has_depth | False |
| seconds | 126.600 |
| matcher | exhaustive |
| registered | 47 |
| requested | 47 |
| n_models | 1 |
| split_warning | False |
| sparse_points | 23,825 |
raw json
{
"backend": "colmap",
"views": 47,
"baseline_recon_units": 9.4358,
"points": 23825,
"has_depth": false,
"seconds": 126.6,
"matcher": "exhaustive",
"registered": 47,
"requested": 47,
"n_models": 1,
"split_warning": false,
"sparse_points": 23825
}20_scale — Metric scale
Converts the scale-free reconstruction to metres using the D455 depth map. This is what replaces Re3Sim's ArUco marker.
How to read it. `spread` is agreement between independent per-frame estimates: <5% consistent, 5-15% loose, >15% do not trust. `baseline_m` should match the real extent of the capture.
| backend | colmap |
|---|---|
| method | sparse_points |
| scale_m_per_unit | 0.284 |
| spread | 0.047 |
| verdict | consistent |
| frames_used | 1 |
| frames_total | 1 |
| baseline_m | 2.683 |
raw json
{
"backend": "colmap",
"method": "sparse_points",
"scale_m_per_unit": 0.284302057399257,
"spread": 0.04707089229306618,
"verdict": "consistent",
"frames_used": 1,
"frames_total": 1,
"baseline_m": 2.682620854988666,
"per_frame": [
{
"scale": 0.284302057399257,
"median_scale": 0.2876123905162095,
"n": 377766,
"rel_mad": 0.04707089229306618,
"trusted": true,
"method": "sparse_points"
}
]
}30_ground — Ground plane / world frame
Fits the table plane and builds a world frame with +Z up and the surface at z=0, so MuJoCo gravity and object placement mean something.
not run
40_mask — Robot mask
SAM3 text-prompted segmentation of the robot, which must be kept out of the background splat (MuJoCo renders it from the URDF).
not run
50_splat — Background splat
Gaussian splat trained on the masked images, initialised from the measured depth cloud rather than sparse SfM points.
How to read it. `eval` is on HELD-OUT views -- the honest number. Training-view PSNR only measures overfitting. `eval.depth.bias_mm` is the early warning for a wrong scale factor.
| backend | gsplat |
|---|---|
| ply | /ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/50_splat/background.ply |
| n_gaussians | 1,000,000 |
| train_seconds | 180.100 |
| peak_vram_gb | 1.680 |
| iters | 7,000 |
| cap_max | 1,000,000 |
| sh_degree | 3 |
| init_points | 499,986 |
| train_views | 41 |
| masked | True |
| depth_weight | 0.500 |
| freespace_weight | 0.100 |
| n_views | 6 |
|---|---|
| lpips | 0.236 |
| psnr | 24.904 |
| ssim | 0.820 |
| bias_mm | -3.250 |
|---|---|
| inlier_frac_10mm | 0.118 |
| mean_abs_mm | 226.016 |
| median_abs_mm | 71.727 |
| n | 178,356.167 |
| p95_abs_mm | 1,061.624 |
| rms_mm | 495.142 |
| valid_frac | 0.697 |
raw json
{
"backend": "gsplat",
"ply": "/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/50_splat/background.ply",
"n_gaussians": 1000000,
"train_seconds": 180.1,
"peak_vram_gb": 1.68,
"eval": {
"n_views": 6,
"lpips": 0.23575166364510855,
"psnr": 24.904295830777595,
"ssim": 0.8196163177490234,
"depth": {
"bias_mm": -3.2504506431091382,
"inlier_frac_10mm": 0.11775478648555233,
"mean_abs_mm": 226.01587483385856,
"median_abs_mm": 71.7267096042633,
"n": 178356.16666666666,
"p95_abs_mm": 1061.6236040989554,
"rms_mm": 495.1416629953069,
"valid_frac": 0.6967037760416667
}
},
"iters": 7000,
"cap_max": 1000000,
"sh_degree": 3,
"init_points": 499986,
"train_views": 41,
"masked": true,
"depth_weight": 0.5,
"freespace_weight": 0.1
}60_mesh — Static geometry
TSDF fusion of the depth maps into a mesh with real measured thickness, plus convex parts for collision.
How to read it. `extent_m` must match the real object. MuJoCo hulls a mesh whole, so the convex parts are what make concave shapes collide right.
| visual_mesh | /ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/static.obj |
|---|---|
| views_fused | 47 |
| label | support |
| voxel_m | 0.006 |
| vertices | 53,929 |
| faces | 100,915 |
raw json
{
"visual_mesh": "/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/static.obj",
"views_fused": 47,
"label": "support",
"seat_slab": {
"centre": [
-0.11397890371008124,
0.21600000000000003,
0.6407894736842105
],
"half_extent": [
0.17891896211005442,
0.189,
0.02
]
},
"voxel_m": 0.006,
"vertices": 53929,
"faces": 100915,
"extent_m": [
0.5265090048542318,
0.5025527209781968,
0.762
],
"collision_parts": [
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_000.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_001.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_002.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_003.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_004.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_005.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_006.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_007.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_008.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_009.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_010.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/60_mesh/collision/part_011.obj"
]
}70_objects — Objects
Per-object mesh and physics. Mass is hand-tuned; inertia follows from the mesh at that mass.
How to read it. `extents_m` is the reality check. A 4 mm slab means one-sided observation (the RoboSnap failure); anything over ~60 cm means the views are not aligning and the poses are too imprecise.
| object | size (cm) | mass | parts | watertight |
|---|---|---|---|---|
| green cup | 8.9 x 9.9 x 18.4 | 0.15 kg | 20 | True |
| small bottle | 4.1 x 4.6 x 9.9 | 0.3 kg | 2 | False |
raw json
{
"objects": [
{
"name": "green cup",
"slug": "green_cup",
"mesh": "/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/mesh.obj",
"collision_parts": [
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull00.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull01.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull02.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull03.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull04.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull05.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull06.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull07.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull08.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull09.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull10.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull11.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull12.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull13.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull14.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull15.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull16.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull17.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull18.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/green_cup/collision/hull19.obj"
],
"position": [
-0.03080668825155623,
0.24856851330194288,
0.7526797142191165
],
"mass_kg": 0.15,
"friction": 1.0,
"com": [
0.003434958095481648,
-0.0038943318169243077,
-0.00634717730837861
],
"inertia": [
[
0.0003447181281606203,
9.74920956245587e-07,
2.4686969554354103e-05
],
[
9.74920956245587e-07,
0.0003517431479083333,
-5.910083414771122e-06
],
[
2.4686969554354103e-05,
-5.910083414771122e-06,
5.5104752736585506e-05
]
],
"extents_m": [
0.08882357213689121,
0.09855806849014094,
0.1840844780768887
],
"watertight": true,
"single_view_extents_m": [
0.08882357213689121,
0.09855806849014093,
0.1840844780768887
],
"merge_inflation": 0.8621416382802122,
"n_points": 305223,
"mode": "recgen_generative",
"source_view": 17
},
{
"name": "small bottle",
"slug": "small_bottle",
"mesh": "/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/small_bottle/mesh.obj",
"collision_parts": [
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/small_bottle/collision/hull00.obj",
"/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/70_objects/small_bottle/collision/hull01.obj"
],
"position": [
-0.18420681626122623,
0.12239262478409875,
0.6961272348965897
],
"mass_kg": 0.3,
"friction": 1.0,
"com": [
0.0221289125207711,
0.0012342997851460024,
0.06953201447952542
],
"inertia": [
[
-0.0015060243067724374,
7.915131349379669e-05,
0.00044312847888503004
],
[
7.915131349379669e-05,
-0.001577941076319601,
5.269313615364336e-05
],
[
0.00044312847888503004,
5.269313615364336e-05,
-0.000143864138258935
]
],
"extents_m": [
0.041315582020125946,
0.045543733639537703,
0.09921177171000928
],
"watertight": false,
"single_view_extents_m": [
0.041315582020125946,
0.045543733639537703,
0.09921177171000928
],
"merge_inflation": 0.9922704264361828,
"n_points": 72913,
"mode": "recgen_generative",
"source_view": 4
}
]
}
80_scene — MuJoCo scene
Assembles the MJCF: static geometry, objects, cameras at the real capture poses, robot.
How to read it. `compiles` must be true. The splat is deliberately absent -- MuJoCo cannot render gaussians; it is composited in stage 90.
| xml | /ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/80_scene/scene.xml |
|---|---|
| bodies | 4 |
| geoms | 16 |
| nq | 14 |
| cameras | 47 |
| objects | 2 |
| collision_parts | 9 |
| compiles | True |
raw json
{
"xml": "/ihub/homedirs/svs_ald/sudhir/real2sim/work/lab47d/80_scene/scene.xml",
"bodies": 4,
"geoms": 16,
"nq": 14,
"cameras": 47,
"objects": 2,
"collision_parts": 9,
"compiles": true
}90_verify — Verification vs real photographs
Renders the sim, composites the splat behind it, and diffs against the actual photograph from that pose. The end-to-end check.
How to read it. This catches the silent failures: wrong scale, rotated background, objects floating. Compare held-out views against training views -- a large gap means the capture was too sparse.
| views | 6 |
|---|---|
| splat | True |
| frame | 20.000 |
|---|---|
| lpips | 0.321 |
| psnr | 19.951 |
| sim_pixel_frac | 0.058 |
| ssim | 0.740 |
raw json
{
"summary": {
"frame": 20.0,
"lpips": 0.3206431542833646,
"psnr": 19.950926324695157,
"sim_pixel_frac": 0.058202473958333334,
"ssim": 0.7399264772733053
},
"per_view": [
{
"psnr": 22.22838088926459,
"ssim": 0.7991556525230408,
"lpips": 0.25035372376441956,
"frame": 0,
"sim_pixel_frac": 0.0533046875
},
{
"psnr": 18.621362931447724,
"ssim": 0.6574383974075317,
"lpips": 0.369310200214386,
"frame": 8,
"sim_pixel_frac": 0.048078125
},
{
"psnr": 19.676231944132013,
"ssim": 0.7342517971992493,
"lpips": 0.3948727548122406,
"frame": 16,
"sim_pixel_frac": 0.06287890625
},
{
"psnr": 21.547707742283304,
"ssim": 0.844158947467804,
"lpips": 0.1992785483598709,
"frame": 24,
"sim_pixel_frac": 0.067234375
},
{
"psnr": 18.096622028043853,
"ssim": 0.6713008284568787,
"lpips": 0.35140836238861084,
"frame": 32,
"sim_pixel_frac": 0.046171875
},
{
"psnr": 19.53525241299945,
"ssim": 0.7332532405853271,
"lpips": 0.3586353361606598,
"frame": 40,
"sim_pixel_frac": 0.071546875
}
],
"views": 6,
"splat": true
}
