/awesome-metric-learning

awesome-metric-learning

  1. In Defense of the Triplet Loss for Person Re-Identification - Proposes and studies various sampling methodologies for triplet selection. Shows evidence batch-hard sampling is effective.
  2. Learning Spread-out Local Features - Attempts to solve mode collapse problem
  3. Combination of Multiple Global Descriptors for Image Retrieval - Respresents image as an ensemble of global descriptors. Joint training on ranking + classification loss.
  4. Visualizing Deep Similarity Networks - Method for visualizing which parts of between two images can be attributable to higher similarity scores.