Closure Classification on Zappos50k in-domain (test)
94.37Top-1 AccMSCon
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MSConLoss=Multi-Similarity Contrastive, Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 94.37 | 0.35 | |
| XEnt MTLoss=Cross-Entropy (Multi-task), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 93.33 | 0.36 | |
| XEnt CloLoss=Cross-Entropy (Closure), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 92.28 | 0.35 | |
| SupCon CloLoss=SupCon (Closure), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 91.75 | 0.41 | |
| SimCLRLoss=SimCLR, Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 81.3 | 0.49 | |
| XEnt CatLoss=Cross-Entropy (Category), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 74.55 | 0.38 | |
| XEnt GendLoss=Cross-Entropy (Gender), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 73.28 | 0.37 | |
| SupCon CatLoss=SupCon (Category), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 73.02 | 0.36 | |
| CSNLoss=Conditional Similarity Network, Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 72.12 | 0.36 | |
| SupCon GenLoss=SupCon (Gender), Evaluation Protocol=Fine-tune classification layer on frozen embedding2023.07 | 69.52 | 0.38 |