Research map / Detection and segmentation / Open-set and few-shot detection
CarFormer: Self-Driving with Learned Object-Centric Representations
Shadi Hamdan, Fatma Guney
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 exact object attributes with learned slot representations | CarFormer · Longest6 | Driving Score: 71.53 -> 74.89 | Improved |
| use object-level slots instead of scene-level VQ-VAE BEV tokens | CarFormer · Longest6 | Driving Score: 17.07 -> 74.89 | Improved |
| predict waypoints with GRU head instead of quantized transformer head (ablation) | CarFormer (slots) · Longest6 | Driving Score: 66.87 -> 74.89 | Improved |
| add block attention (ablation) | CarFormer (slots) · Longest6 | Driving Score: 70.42 -> 74.89 | Improved |
| add auxiliary slot forecasting (ablation) | CarFormer (slots) · Longest6 | Driving Score: 57.25 -> 74.89 | Improved |
| add creeping during evaluation (ablation) | CarFormer (slots) · Longest6 | Driving Score: 69.16 -> 74.89 | Improved |
| remove creeping (infraction score) (ablation) | CarFormer (slots) · Longest6 | Infraction Score: 0.79 -> 0.83 | Improved |
| increase number of slots from 7 to 30 (ablation) | CarFormer (slots, not enlarged) · Longest6 | Driving Score: 48.17 -> 71.48 | Improved |
| enlarge small vehicles in BEV for slot extraction (ablation) | CarFormer (slots, 7 slots) · Longest6 | Driving Score: 48.17 -> 62.93 | Improved |
| enlarge small vehicles in BEV for slot extraction (ablation) | CarFormer (slots, 30 slots) · Longest6 | Driving Score: 71.48 -> 74.89 | Improved |
| CarFormer with exact attributes vs PlanT architecture | CarFormer (attributes) · Longest6 | Driving Score: 73.36 -> 71.53 | No clear effect |
Closest papers by meaning
- BEVFormer: Learning Bird’s-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformers (ECCV 2022 · similarity 0.89)
- BEVFormer: Learning Bird’s-Eye-View Representation From LiDAR-Camera via Spatiotemporal Transformers (TPAMI 2024 · similarity 0.88)
- Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline (TPAMI 2024 · similarity 0.88)
- Holistic Autonomous Driving Understanding by Bird's-Eye-View Injected Multi-Modal Large Models (CVPR 2024 · similarity 0.86)
- Transfer Your Perspective: Controllable 3D Generation from Any Viewpoint in a Driving Scene (CVPR 2025 · similarity 0.86)
- Leveraging Driver Field-of-View for Multimodal Ego-Trajectory Prediction (ICLR 2025 · similarity 0.86)
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.
