Research map / Architectures and efficiency / Attention and transformers
V2X-ViT: Vehicle-to-Everything Cooperative Perception with Vision Transformer
Runsheng Xu, Hao Xiang, Zhengzhong Tu, Xin Xia, Ming-Hsuan Yang, Jiaqi Ma
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 multi-scale window attention (MSwin) (ablation) | V2X-ViT base · V2XSet Noisy | AP@0.7: 0.478 -> 0.519 | Improved |
| add split attention fusion of window branches (ablation) | V2X-ViT base + MSwin · V2XSet Noisy | AP@0.7: 0.519 -> 0.548 | Improved |
| add heterogeneous multi-agent self-attention (HMSA) (ablation) | V2X-ViT base + MSwin + SpAttn · V2XSet Noisy | AP@0.7: 0.548 -> 0.601 | Improved |
| add delay-aware positional encoding (DPE) (ablation) | V2X-ViT · V2XSet Noisy | AP@0.7: 0.601 -> 0.614 | Improved |
| add DPE under 100 ms delay (ablation) | V2X-ViT · V2XSet Noisy (100 ms delay) | AP@0.7: 0.639 -> 0.650 | Improved |
| add DPE under 400 ms delay (ablation) | V2X-ViT · V2XSet Noisy (400 ms delay) | AP@0.7: 0.458 -> 0.478 | Improved |
| use 3 encoder layers instead of 1 (ablation) | V2X-ViT · V2XSet Noisy | AP@0.7: 0.591 -> 0.614 | Improved |
| transform features to ego range before cropping instead of broadcasting (ablation) | V2X-ViT · V2XSet Perfect | AP@0.7: 0.623 -> 0.712 | Improved |
| transform features to ego range before cropping instead of broadcasting (ablation) | DiscoNet · V2XSet Perfect | AP@0.7: 0.610 -> 0.695 | Improved |
| add noise (pose error and 100 ms delay) to evaluation | Early Fusion · V2XSet | AP@0.7: 0.710 -> 0.384 | Hurt |
| add noise (pose error and 100 ms delay) to evaluation | Late Fusion · V2XSet | AP@0.7: 0.620 -> 0.307 | Hurt |
| add noise (pose error and 100 ms delay) to evaluation | V2X-ViT · V2XSet | AP@0.7: 0.712 -> 0.614 | Hurt |
Closest papers by meaning
- Privacy-Preserving V2X Collaborative Perception Integrating Unknown Collaborators (AAAI 2025 · similarity 0.89)
- End-to-End Autonomous Driving Through V2X Cooperation (AAAI 2025 · similarity 0.87)
- V2X-Radar: A Multi-modal Dataset with 4D Radar for Cooperative Perception (NeurIPS 2025 · similarity 0.87)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.86)
- BEVSync: Asynchronous Data Alignment for Camera-based Vehicle-Infrastructure Cooperative Perception Under Uncertain Delays (AAAI 2025 · similarity 0.86)
- UrbanIng-V2X: A Large-Scale Multi-Vehicle, Multi-Infrastructure Dataset Across Multiple Intersections for Cooperative Perception (NeurIPS 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.
