Initial dataset: PerceptPick benchmark assets in canonical tree layout

.gitattributes

@@ -0,0 +1,14 @@
+# Mesh files
+*.obj filter=lfs diff=lfs merge=lfs -text
+*.ply filter=lfs diff=lfs merge=lfs -text
+*.stl filter=lfs diff=lfs merge=lfs -text
+*.dae filter=lfs diff=lfs merge=lfs -text
+*.urdf filter=lfs diff=lfs merge=lfs -text
+*.mtl filter=lfs diff=lfs merge=lfs -text
+# Image textures
+*.png filter=lfs diff=lfs merge=lfs -text
+*.jpg filter=lfs diff=lfs merge=lfs -text
+*.jpeg filter=lfs diff=lfs merge=lfs -text
+*.bmp filter=lfs diff=lfs merge=lfs -text
+*.tiff filter=lfs diff=lfs merge=lfs -text
+*.gif filter=lfs diff=lfs merge=lfs -text

ASSET_BUNDLE.md

@@ -0,0 +1,78 @@
+# PerceptPick — pre-prepared assets
+
+This bundle holds the URDF / VHACD / mesh assets for the YCB-V dataset
+across nine mesh sources (oracle CAD plus eight reconstruction methods),
+together with FoundationPose and MegaPose pose-estimator CSVs.
+
+Drop into a `perceptpick` clone to skip Stage A (`01_prepare_assets.py`)
+and the FoundationPose / MegaPose pipelines:
+
+```
+git clone <perceptpick>
+cd perceptpick
+
+# 1. download the BOP YCBV test split (scenes 48-59 + models)
+#    see the README's "Get the YCB-Video dataset" section.
+
+# 2. unpack this bundle next to the repo
+unzip perceptpick_assets.zip
+
+# 3. wire the bundle into the expected path
+mkdir -p assets
+mv perceptpick_assets/ycbv assets/ycbv
+```
+
+After that, jump straight to Stage B / C — no Stage A re-prep needed.
+
+## Layout
+
+```
+ycbv/
+├── GT/                                          # oracle CAD (BOP YCBV models)
+│   ├── meshes/obj_NNNNNN.{obj,mtl,png}
+│   ├── vhacd/obj_NNNNNN_vhacd.obj
+│   ├── urdf/obj_NNNNNN.urdf
+│   └── pose_estimates/
+│       ├── FoundationPose.csv                   # FoundationPose on GT meshes
+│       └── MegaPose.csv                         # MegaPose on GT meshes
+├── BakedSDF/                                    # 8 reconstruction methods
+│   ├── meshes/, vhacd/, urdf/
+│   └── pose_estimates/
+│       ├── FoundationPose.csv                   # FoundationPose on BakedSDF
+│       └── MegaPose.csv                         # MegaPose on BakedSDF
+├── MonoSDF/, Nerfacto/, Neuralangelo/
+├── NGP/, RealCAP/, UniSurf/, VolSDF/
+```
+
+Each method folder is fully self-contained: the meshes the simulator
+loads, the URDFs and VHACDs the physics layer needs, and the pose CSVs
+that were generated using *that* mesh as the pose-estimator's reference
+model. The CSVs are tiny; the meshes / VHACDs make up almost all of the
+disk footprint.
+
+## URDF paths
+
+URDFs reference the sibling collision mesh with a relative path:
+`<mesh filename="../vhacd/obj_NNNNNN_vhacd.obj"/>`. No absolute paths,
+no system-specific roots — the bundle is portable.
+
+## Running the benchmark
+
+```bash
+# Stage B — sample antipodal grasps + simulate, per (object, gripper) on the GT meshes
+pixi run python scripts/02_grasp_sweep.py --dataset ycbv --mesh-source GT --n-grasps 5000
+
+# Stage C, Condition 1 — Oracle / Oracle (ideal baseline)
+pixi run python scripts/04_evaluate.py --dataset ycbv \
+    --gt-mesh GT --est-mesh GT \
+    --pose-csv FoundationPose.csv --gripper auto --workers 4 --resume --headless
+
+# Stage C, Condition 3 — End-to-end realistic (BakedSDF mesh + BakedSDF-conditioned pose)
+pixi run python scripts/04_evaluate.py --dataset ycbv \
+    --gt-mesh BakedSDF --est-mesh BakedSDF \
+    --pose-csv FoundationPose.csv --gripper auto --workers 4 --resume --headless
+```
+
+If you'd rather regenerate the assets from scratch (e.g. to verify VHACD
+parameters), ignore this bundle and run
+`scripts/01_prepare_assets.py --dataset ycbv --all-mesh-sources`.

README.md

@@ -0,0 +1,131 @@
+---
+license: cc-by-4.0
+task_categories:
+  - robotics
+tags:
+  - robotics
+  - grasping
+  - pose-estimation
+  - 3d-reconstruction
+  - simulation
+  - ycb
+  - icra
+pretty_name: PerceptPick Benchmark
+size_categories:
+  - 1G<n<10G
+---
+
+# PerceptPick — Benchmarking the Effects of Object Pose Estimation and Reconstruction on Robotic Grasping Success
+
+This dataset accompanies the paper accepted at **IEEE ICRA 2026**:
+
+> **Benchmarking the Effects of Object Pose Estimation and Reconstruction on Robotic Grasping Success**
+> Varun Burde, Pavel Burget, Torsten Sattler — Czech Technical University in Prague.
+> [arXiv:2602.17101](https://arxiv.org/abs/2602.17101) · [Project page](https://varunburde.github.io/perceptpick) · [Code](https://github.com/varunburde/perceptpick)
+
+## Abstract
+
+3D reconstruction serves as the foundational layer for numerous robotic perception tasks, including 6D object pose estimation and grasp pose generation. Modern 3D reconstruction methods for objects can produce visually and geometrically impressive meshes from multi-view images, yet standard geometric evaluations do not reflect how reconstruction quality influences downstream tasks such as robotic manipulation performance. This paper addresses this gap by introducing a large-scale, physics-based benchmark that evaluates 6D pose estimators and 3D mesh models based on their functional efficacy in grasping. We analyze the impact of model fidelity by generating grasps on various reconstructed 3D meshes and executing them on the ground-truth model, simulating how grasp poses generated with an imperfect model affect interaction with the real object. This assesses the combined impact of pose error, grasp robustness, and geometric inaccuracies from 3D reconstruction. Our results show that reconstruction artifacts significantly decrease the number of grasp pose candidates but have a negligible effect on grasping performance given an accurately estimated pose. Our results also reveal that the relationship between grasp success and pose error is dominated by spatial error, and even a simple translation error provides insight into the success of the grasping pose of symmetric objects.
+
+---
+
+## Layout
+
+The dataset ships in the canonical tree layout used by the PerceptPick code — one folder per **mesh source** (oracle CAD `GT` plus eight reconstruction methods), each containing pre-built meshes, V-HACD convex decompositions, URDFs, and BOP-style pose-estimator CSVs.
+
+```
+ycbv/
+├── GT/                                # Oracle CAD baseline
+│   ├── meshes/obj_NNNNNN.obj
+│   ├── vhacd/obj_NNNNNN_vhacd.obj
+│   ├── urdf/obj_NNNNNN.urdf
+│   └── pose_estimates/
+│       ├── FoundationPose.csv         # FoundationPose run on GT meshes
+│       └── MegaPose.csv               # MegaPose run on GT meshes
+├── BakedSDF/, MonoSDF/, Nerfacto/     # Reconstructed-mesh sources
+│   ├── meshes/, vhacd/, urdf/
+│   └── pose_estimates/
+│       ├── FoundationPose.csv         # FoundationPose run on this method's meshes
+│       └── MegaPose.csv
+├── Neuralangelo/, NGP/, RealCAP/, UniSurf/, VolSDF/
+```
+
+The tree maps 1-to-1 onto the `assets/ycbv/` folder of the [code repo](https://github.com/varunburde/perceptpick).
+
+## Reconstruction methods
+
+| Method | Type | Description |
+|---|---|---|
+| **GT** | Oracle | Ground-truth CAD model (upper bound) |
+| **BakedSDF** | Neural SDF | Baked signed distance field reconstruction |
+| **MonoSDF** | Neural SDF | Monocular depth-guided SDF |
+| **Nerfacto** | NeRF | Nerfstudio's factored NeRF |
+| **Neuralangelo** | Neural SDF | High-fidelity surface reconstruction from video |
+| **NGP** | NeRF | Instant Neural Graphics Primitives |
+| **RealCAP** | Classical | Real-world capture baseline |
+| **UniSurf** | Neural SDF | Unified surface reconstruction |
+| **VolSDF** | Neural SDF | Volume rendering with SDF |
+
+## Pose estimators
+
+- **FoundationPose** — transformer-based 6-DoF pose estimator
+- **MegaPose** — large-scale generalizable pose estimator
+
+Each `pose_estimates/<Estimator>.csv` is a BOP-style submission CSV with columns
+`scene_id, im_id, obj_id, score, R, t, time`. The estimator was fed the mesh source
+of its parent folder at inference time (e.g.
+`ycbv/BakedSDF/pose_estimates/FoundationPose.csv` is FoundationPose run on the
+BakedSDF reconstructed meshes).
+
+## Objects
+
+21 YCB objects from the YCB-Video object set (object IDs 1–21).
+
+## Download
+
+```bash
+# Pull the entire tree directly into ./assets (~5.6 GB)
+hf download varunburde/perceptpick --repo-type=dataset --local-dir assets
+```
+
+Or use the legacy CLI:
+
+```bash
+huggingface-cli download varunburde/perceptpick --repo-type=dataset --local-dir assets
+```
+
+To consume the data from Python:
+
+```python
+import pandas as pd
+from huggingface_hub import hf_hub_download
+
+csv_path = hf_hub_download(
+    "varunburde/perceptpick",
+    "ycbv/BakedSDF/pose_estimates/FoundationPose.csv",
+    repo_type="dataset",
+)
+df = pd.read_csv(csv_path)
+print(df.head())
+```
+
+## Citation
+
+```bibtex
+@inproceedings{burde2026perceptpick,
+  title     = {Benchmarking the Effects of Object Pose Estimation and
+               Reconstruction on Robotic Grasping Success},
+  author    = {Burde, Varun and Burget, Pavel and Sattler, Torsten},
+  booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
+  year      = {2026},
+  eprint    = {2602.17101},
+  archivePrefix = {arXiv},
+  primaryClass  = {cs.RO},
+  doi       = {10.48550/arXiv.2602.17101}
+}
+```
+
+## License
+
+This dataset is released under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/).
+The YCB-Video subset follows the original [YCB dataset license](http://www.ycbbenchmarks.com/).

ycbv/BakedSDF/meshes/obj_000001.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000001.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f176baf768337b2553f3415c84f63b12250c329e04e84364004f9af50eaedff
+size 6322379

ycbv/BakedSDF/meshes/obj_000001.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000002.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000002.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac5a104b329c34767c64f6892bcfcbb621bc657e93e42f46d3fe751fba4dc322
+size 6154206

ycbv/BakedSDF/meshes/obj_000002.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000003.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000003.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3bb96bf72bff8c16d858b7869d7741cc39f8c535a357d548c04389f760ce346
+size 6284516

ycbv/BakedSDF/meshes/obj_000003.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000004.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000004.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4e87b0f89eb249fa09be732fdd10c3c623ae99cfd97c620c0bd6568df4f3b42
+size 4400930

ycbv/BakedSDF/meshes/obj_000004.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000005.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000005.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e73a1367aa8451cf6c4ceaa8d9abe3e216a485d40b28788f51b31449140ed90
+size 6990610

ycbv/BakedSDF/meshes/obj_000005.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000006.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000006.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c50331de59ed1fded31a123f405f7e807a244acd5d214a2ad083474ed5fcbcaa
+size 6431789

ycbv/BakedSDF/meshes/obj_000006.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000007.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000007.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8074ea0309277abd7270b6d18b7b647d7a9501125b0463126eec4a34b49ae50d
+size 6282242

ycbv/BakedSDF/meshes/obj_000007.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000008.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000008.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1fa43e43f1baf726d952ccaa3db644529ebe5725902d8d095dcca03f7964063
+size 6292276

ycbv/BakedSDF/meshes/obj_000008.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000009.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000009.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecbdb8b7b112f594256cc6a3aac91d1632d72e7e10d5603be6c29622574742a1
+size 4927290

ycbv/BakedSDF/meshes/obj_000009.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000010.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000010.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:679fd0785e1fab54f9b2eeeeedd1fe57827791636a5c13eaee56315f7e29ec7c
+size 6185875

ycbv/BakedSDF/meshes/obj_000010.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000011.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000011.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b1c93dc1fb41c8db331bef835763004eda064862485ad533aaf9682252f95d4
+size 6841420

ycbv/BakedSDF/meshes/obj_000011.obj.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cece129d4d80fed29e3626b36f84f763f3f2bf8d49044306eb7f2804a4e9f705
+size 219

ycbv/BakedSDF/meshes/obj_000012.mtl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f09fc7c6413f577c1533e637066d5ff0d70394fc46b089a4da857eb0ad51bc5d
+size 136

ycbv/BakedSDF/meshes/obj_000012.obj

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dd662368ce9b99c24fdd26b94c38bf26ed22cc6a79cb8536e2189f0f1fe0626
+size 6452185