hominot
commented
6 years ago - Datasets to use
- LFW
- Stanford Online Product
- Models to compare
- Contrastive loss
- Triplet loss
- N-tuplet loss
- Metrics to evaluate
- Recall@k
- Experiment procedure
- Cross validation (?)
- Train/Validation/Test split ratio
- Hyper parameter tuning