Confusezius/Deep-Metric-Learning-Baselines
A PyTorch pipeline implementing deep metric learning methods including triplet loss, margin loss, and proxy-based losses for image similarity tasks.
Velocity · 7d
+0.2
★ / day
Trend
→steady
star history
This repository provides an extendable PyTorch framework for deep metric learning, implementing various loss functions (Triplet, Margin, ProxyNCA, N-Pair) and sampling strategies (random, softhard, semihard, distance). It includes dataloaders for standard benchmark datasets (CUB200, CARS196, Stanford Online Product) used to evaluate image retrieval and similarity learning models.