Research map / Label-efficient and robust learning / Knowledge distillation
VehicleMAE: View-asymmetry Mutual Learning for Vehicle Re-identification Pre-training via Masked AutoEncoders
Qi Wang, Zeyu Zhang, Dong Wang, Di Gai, Xin Xiong, Jiyang Xu, Ruihua Zhou
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 historical distillation (Pa) to baseline MAE (ablation) | VehicleMAE baseline · VeRi-776 unsupervised | mAP: 37.6 -> 40.1 | Improved |
| add present distillation (Pr) to baseline MAE (ablation) | VehicleMAE baseline · VeRi-776 unsupervised | mAP: 37.6 -> 40.5 | Improved |
| add all modules (PPMD and VMIM) to baseline MAE (ablation) | VehicleMAE · VeRi-776 unsupervised | mAP: 37.6 -> 42.5 | Improved |
| add all modules (PPMD and VMIM) to baseline MAE (ablation) | VehicleMAE · VehicleID large unsupervised | Rank-1: 69.0 -> 73.3 | Improved |
| use asymmetric instead of random mask sampling at 75% mask ratio (ablation) | VehicleMAE · VeRi-776 unsupervised | mAP: 40.1 -> 42.5 | Improved |
| use asymmetric instead of random mask sampling at 25% mask ratio (ablation) | VehicleMAE · VeRi-776 unsupervised | mAP: 38.8 -> 40.6 | Improved |
| use asymmetric instead of random mask sampling at 50% mask ratio (ablation) | VehicleMAE · VeRi-776 unsupervised | mAP: 39.2 -> 41.7 | Improved |
| raise asymmetric mask ratio from 25% to 75% (ablation) | VehicleMAE · VeRi-776 unsupervised | mAP: 40.6 -> 42.5 | Improved |
| combine historical and present distillation vs independent MAEs (mean encoder) (ablation) | VehicleMAE · VeRi-776 unsupervised | Rank-1: 84.6 -> 87.6 | Improved |
| combine historical and present distillation vs independent MAEs (mean encoder) (ablation) | VehicleMAE · VehicleID unsupervised | Rank-1: 70.9 -> 73.3 | Improved |
| average two encoders into a mean encoder instead of using one encoder (ablation) | VehicleMAE · VeRi-776 unsupervised | Rank-1: 84.3 -> 84.6 | Improved |
| pre-train with VehicleMAE on DiffVERI instead of CMAE | CLIP-B/16 · VeRi-776 unsupervised | mAP: 41.6 -> 42.5 | Improved |
Closest papers by meaning
- Structural Information Guided Multimodal Pre-training for Vehicle-Centric Perception (AAAI 2024 · similarity 0.89)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.87)
- Infer the Whole from a Glimpse of a Part: Keypoint-Based Knowledge Graph for Vehicle Re-Identification (AAAI 2025 · similarity 0.87)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.86)
- CLIP-driven View-aware Prompt Learning for Unsupervised Vehicle Re-identification (AAAI 2025 · similarity 0.86)
- Asymmetric Masked Distillation for Pre-Training Small Foundation Models (CVPR 2024 · 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.
