Research map / Label-efficient and robust learning / Recognition and long-tail
A2GC: Asymmetric Aggregation with Geometric Constraints for Locally Aggregated Descriptors
Zhenyu Li, Tianyi Shang
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 asymmetric aggregation (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 93.9 -> 94.9 | Improved |
| add geometric constraints (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 94.1 -> 94.9 | Improved |
| add both asymmetric aggregation and geometric constraints (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 92.5 -> 94.9 | Improved |
| fine-tune last 4 blocks instead of frozen backbone (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 92.9 -> 94.9 | Improved |
| fine-tune all blocks instead of last 4 (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 94.9 -> 94.0 | Hurt |
| scale backbone from ViT-S to ViT-g (ablation) | A2GC · Pitts30k | Recall@1: 94.0 -> 95.6 | Improved |
| increase descriptor size from 512+32 to 32768+512 (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 93.7 -> 95.0 | Improved |
| increase descriptor size to 8192+256 (ablation) | A2GC (ViT-B) · Pitts30k | Recall@5: 98.7 -> 98.5 | No clear effect |
| increase input resolution from 112 to 588 (ablation) | A2GC (ViT-B) · MSLS-val | Recall@1: 81.6 -> 96.0 | Improved |
| increase input resolution from 112 to 588 (ablation) | A2GC (ViT-B) · Pitts30k | Recall@1: 92.1 -> 96.7 | Improved |
| A2GC vs SALAD on cross-domain test | A2GC · Nordland (spring vs winter) | Recall@1: 25.18 -> 34.71 | Improved |
| A2GC vs Pair-VPR on MSLS-challenge | A2GC · MSLS-challenge | Recall@1: 81.7 vs 80.6 | Hurt |
Closest papers by meaning
- Optimal Transport Aggregation for Visual Place Recognition (CVPR 2024 · similarity 0.91)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.90)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.89)
- SuperVLAD: Compact and Robust Image Descriptors for Visual Place Recognition (NeurIPS 2024 · similarity 0.88)
- MutualVPR: A Mutual Learning Framework for Resolving Supervision Inconsistencies via Adaptive Clustering (NeurIPS 2025 · similarity 0.88)
- Towards Seamless Adaptation of Pre-trained Models for Visual Place Recognition (ICLR 2024 · similarity 0.87)
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.
