Manolis Savva, Ali Mahdavi-Amiri, Jiayi Liu
We lifted 9 functions out of this paper's own repositories and ran 7 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| 3dlg-hcvc/paris | canonical | 7 of 9 |
| Function | Status | Where it lives |
|---|---|---|
| cast_rays | Ran | 3dlg-hcvc/paris/models/ray_utils.py code served (permissive licence) · get_code("45090c68bfc689c0") |
| dot | Ran | 3dlg-hcvc/paris/models/utils.py code served (permissive licence) · get_code("456eeade4275ae9d") |
| get_activation | Ran | 3dlg-hcvc/paris/models/utils.py code served (permissive licence) · get_code("d39816e07771a285") |
| get_ray_directions | Ran | 3dlg-hcvc/paris/models/ray_utils.py code served (permissive licence) · get_code("436816f5ff880b73") |
| get_rotation_axis_angle | Ran | 3dlg-hcvc/paris/preprocess.py code served (permissive licence) · get_code("4e5f1556e15a1cec") |
| merge_meshsets | Ran | 3dlg-hcvc/paris/preprocess.py code served (permissive licence) · get_code("f01de641f2fc93da") |
| normalize | Ran | 3dlg-hcvc/paris/preprocess.py code served (permissive licence) · get_code("5061e5677030ce23") |
| chunk_batch | Not yet run | 3dlg-hcvc/paris/models/utils.py code served (permissive licence) · get_code("84eb8261612388be") |
| get_rays | Not yet run | 3dlg-hcvc/paris/models/ray_utils.py code served (permissive licence) · get_code("760ae16368cc7e68") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
We address the task of simultaneous part-level reconstruction and motion parameter estimation for articulated objects. Given two sets of multi-view images of an object in two static articulation states, we decouple the movable part from the static part and reconstruct shape and appearance while predicting the motion parameters. To tackle this problem, we present PARIS: a self-supervised, end-to-end architecture that learns part-level implicit shape and appearance models and optimizes motion parameters jointly without any 3D supervision, motion, or semantic annotation. Our experiments show that our method generalizes better across object categories, and outperforms baselines and prior work that are given 3D point clouds as input. Our approach improves reconstruction relative to state-ofthe-art baselines with a Chamfer-L1 distance reduction of 3.94 (45.2%) for objects and 26.79 (84.5%) for parts, and achieves 5% error rate for motion estimation across 10 object categories.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2308.07391")
get_code_for_paper("2308.07391")
have("2308.07391")
Connect an agent — have() is free.