Research map / Detection and segmentation / 3D detection and driving
Super Sparse 3D Object Detection
Lue Fan, Yuxue Yang, Wang Feng, Naiyan Wang, Zhaoxiang Zhang
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 |
|---|---|---|---|
| add instance grouping without SIR (ablation) | FSDplain · Waymo val vehicle | L2 3D APH: 62.29 -> 63.13 | Improved |
| add SIR on top of grouping (ablation) | FSDagg · Waymo val vehicle | L2 3D APH: 63.13 -> 65.20 | Improved |
| add group correction (ablation) | FSDnogc · Waymo val vehicle | L2 3D APH: 65.20 -> 69.30 | Improved |
| use SIR instead of feature diffusion for large vehicles | SSTcenter · Waymo val vehicles 12 m or longer | vehicle AP: 30.5 -> 47.9 | Improved |
| add random skeleton point sampling (ablation) | FSD++ · Waymo val | mean AP/APH: 75.23/73.91 -> 76.10/74.73 | Improved |
| add random skeleton point sampling (ablation) | FSD++ · Waymo val | latency (ms): 73.9 -> 68.7 | Improved |
| increase number of frames from 2 to 6 (ablation) | FSD++ · Waymo val | mean AP/APH: 73.39/71.83 -> 76.10/74.73 | Improved |
| use FSD++ residual point probing instead of FSD point concatenation (6 frames) | FSD · Waymo val | latency (ms): 116.2 -> 68.7 | Improved |
| raise max age from 1 to 2 (ablation) | FSD++ · Waymo val | mean AP/APH: 75.17/73.82 -> 76.10/74.73 | Improved |
| raise max age from 2 to 3 (ablation) | FSD++ · Waymo val | mean AP/APH: 76.10/74.73 -> 76.14/74.74 | No clear effect |
| insert keyframes at regular intervals (ablation) | FSD++ · Waymo val | mean AP/APH: 76.10/74.73 -> 76.05/74.67 | No clear effect |
| drop all initial seed boxes (ablation) | FSD++ · Waymo val | mean AP/APH: 76.10/74.73 -> 74.69/73.35 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- AFDetV2 — AFDetV2: Rethinking the Necessity of the Second Stage for Object Detection from Point Clouds (AAAI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Fully Sparse 3D Object Detection (NeurIPS 2022 · similarity 0.97)
- FSD V2: Improving Fully Sparse 3D Object Detection With Virtual Voxels (TPAMI 2024 · similarity 0.92)
- Fully Sparse Fusion for 3D Object Detection (TPAMI 2024 · similarity 0.91)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.91)
- MsSVT: Mixed-scale Sparse Voxel Transformer for 3D Object Detection on Point Clouds (NeurIPS 2022 · similarity 0.90)
- Far3D: Expanding the Horizon for Surround-View 3D Object Detection (AAAI 2024 · similarity 0.88)
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.
