Research map / Label-efficient and robust learning / Self-supervised and contrastive
The Devil is in the Details: Self-Supervised Attention for Vehicle Re-Identification
Pirazh Khorramshahi, Neehar Peri, Jun-cheng Chen, Rama Chellappa
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 self-supervised residual attention (SAVER) | global appearance baseline · VeRi | mAP (%): 78.2 -> 79.6 | Improved |
| add SAVER | baseline · VehicleID | CMC@1 (%): 78.4 -> 79.9 | Improved |
| add SAVER | baseline · VERI-Wild | mAP (%): 65.0 -> 67.7 | Improved |
| add SAVER | baseline · Vehicle-1M | CMC@1 (%): 93.6 -> 95.5 | Improved |
| add SAVER | baseline · VD1 | mAP (%): 96.4 -> 96.7 | No clear effect |
| add SAVER | baseline · VD2 | mAP (%): 96.8 -> 96.7 | No clear effect |
| use VAE instead of AE for residual generation (ablation) | SAVER · VeRi | mAP (%): 79.0 -> 79.6 | Improved |
| use VAE instead of GAN for residual generation (ablation) | SAVER · VeRi | mAP (%): 78.3 -> 79.6 | Improved |
| use VAE instead of bilateral filtering for residual generation (ablation) | SAVER · VeRi | mAP (%): 78.5 -> 79.6 | Improved |
| use convex combination of image and residual instead of residual only (ablation) | SAVER · VeRi | mAP (%): 77.5 -> 79.6 | Improved |
| use convex combination instead of concatenating residual as extra channels (ablation) | SAVER · VeRi | mAP (%): 75.7 -> 79.6 | Improved |
| use convex combination instead of point-wise excitation by residual (ablation) | SAVER · VeRi | mAP (%): 71.4 -> 79.6 | Improved |
Closest papers by meaning
- Self-Supervised Geometric Features Discovery via Interpretable Attention for Vehicle Re-Identification and Beyond (ICCV 2021 · similarity 0.90)
- Learning Coarse-to-Fine Structured Feature Embedding for Vehicle Re-Identification (AAAI 2018 · similarity 0.89)
- CLIP-driven View-aware Prompt Learning for Unsupervised Vehicle Re-identification (AAAI 2025 · similarity 0.87)
- Infer the Whole from a Glimpse of a Part: Keypoint-Based Knowledge Graph for Vehicle Re-Identification (AAAI 2025 · similarity 0.86)
- Alice Benchmarks: Connecting Real World Re-Identification with the Synthetic (ICLR 2024 · similarity 0.85)
- Structural Information Guided Multimodal Pre-training for Vehicle-Centric Perception (AAAI 2024 · similarity 0.85)
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.
