- GitHub - pavelsuma/ames: AMES: Asymmetric and Memory-Efficient Similarity
- [2412.00784] EDTformer: An Efficient Decoder Transformer for Visual Place Recognition
- [2406.16204] Breaking the Frame: Visual Place Recognition by Overlap Prediction
Local Feature Based
VLAD
Deep Learning Based
DELF
DELG
NetVLAD
MAC
GeM
Losses / Objectives
Datasets / Benchmarks
Oxford5K
ROxford5K
Google Landmarks
MET
Products-10K
Large scale product recognition dataset
AICrowd Products
AIcrowd | Visual Product Recognition Challenge 2023 | Challenges
Amazon Reviews
Tricks of the Trade
Query Expansion
Index Augmentation
Reranking
Spatial Verification
Ideas
- tokenformer for instance recognition (learned codebook)
- VLMs for instance recognition (same or not, prompting with multiple examples ⇒ generate query token (prefixLM style masking))