Research map / Detection and segmentation / 3D detection and driving
OmniNOCS: A unified NOCS dataset and model for 3D lifting of 2D objects
Akshay Krishnan, Abhijit Kundu, Kevis-Kokitsi Maninis, James Hays, Matthew Brown
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace MaskRCNN conv NOCS head with transformer head (ablation) | NOCSformer · OmniNOCS | NOCS PSNR: +5.23dB | Improved |
| replace MaskRCNN conv NOCS head with transformer head (ablation) | NOCSformer · OmniNOCS | Mask IoU: +21.27% | Improved |
| replace Hourglass head with transformer NOCS head (ablation) | NOCSformer · OmniNOCS | NOCS PSNR: +3.06dB | Improved |
| use discretized (binned classification) NOCS prediction instead of continuous regression (ablation) | NOCSformer · OmniNOCS | NOCS PSNR: +4.11 dB | Improved |
| use NOCS-based localization instead of box regression for zero-shot transfer | NOCSformer vs Cube R-CNN · NOCS-Real275 (unseen) | mAP 3D IoU@25: 14.9 -> 43.5 | Improved |
| supervise orientation directly with canonical labels instead of Chamfer loss | 3D orientation heads · KITTI-val (heading axis) | accuracy @5 deg: 19.43 -> 41.99 | Improved |
| predict orientation from NOCS-based model instead of Cubeformer box regression | NOCSformer · KITTI-val (heading axis) | accuracy @5 deg: 41.99 -> 49.16 | Improved |
| multi-dataset NOCSformer vs single-dataset specialists on nuScenes | NOCSformer · nuScenes | mATE / mAOE: 0.887 vs 0.653; 0.558 vs 0.319 | Hurt |
| zero-shot NOCSformer vs supervised NOCS model on Real275 | NOCSformer · NOCS-Real275 | mAP 3D IoU@50: 10.6 vs 72.4 | Hurt |
Closest papers by meaning
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.87)
- LabelAny3D: Label Any Object 3D in the Wild (NeurIPS 2025 · similarity 0.86)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.86)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.86)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.86)
- MonoMAE: Enhancing Monocular 3D Detection through Depth-Aware Masked Autoencoders (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
