Research map / Detection and segmentation / 3D detection and driving
Point2Seq: Detecting 3D Objects As Sequences
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the anchor-based head with Point2Seq | SECOND on Waymo Open Dataset · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 73.62 -> 77.52 | Helped |
| Replace the center-based head with Point2Seq | CenterPoints on Waymo Open Dataset · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 75.58 -> 77.52 | Helped |
| Use Point2Seq instead of the center-based detector | Voxel-based 3D detection · ONCE validation, overall | mAP (%): 60.05 -> 66.16 | Helped |
| Add similarity-based matching to the convolutional head (ablation) | Voxel-based convolutional detector · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 73.62 -> 76.72 | Helped |
| Add the scene-to-sequence decoder to similarity-based matching (ablation) | Voxel-based detector with similarity-based matching · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 76.72 -> 77.52 | Helped |
| Use a pillar backbone with Point2Seq instead of an anchor head (ablation) | Pillar-based detector · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 63.31 -> 69.01 | Helped |
| Use a voxel backbone with Point2Seq instead of a center head (ablation) | Voxel-based detector · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 75.58 -> 77.52 | Helped |
| Move orientation after size in the word sequence (ablation) | Point2Seq · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 77.52 -> 73.82 | Hurt |
| Move category from the end to the beginning of the attribute words (ablation) | Point2Seq · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 77.52 -> 75.96 | Hurt |
| Use corner-coordinate differences instead of the proposed sequence-distance term (ablation) | Point2Seq similarity matching · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 75.40 -> 74.31 | Hurt |
| Use 3D IoU instead of the proposed sequence-distance term (ablation) | Point2Seq similarity matching · Waymo Open Dataset validation, Vehicle LEVEL 1 | mAP (%): 75.40 -> 75.07 | Hurt |
Closest papers by meaning
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.92)
- 3D-AVS: LiDAR-based 3D Auto-Vocabulary Segmentation (CVPR 2025 · similarity 0.89)
- Towards 3D Objectness Learning in an Open World (NeurIPS 2025 · similarity 0.88)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.88)
- OpenGaussian: Towards Point-Level 3D Gaussian-based Open Vocabulary Understanding (NeurIPS 2024 · similarity 0.88)
- SP3D: Boosting Sparsely-Supervised 3D Object Detection via Accurate Cross-Modal Semantic Prompts (CVPR 2025 · 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.