Feature Upsampling On Imagenet
Metrics
Average Drop
Average Increase
Results
Performance results of various models on this benchmark
| Paper Title | |||
|---|---|---|---|
| Strided | 11.48 | 4.97 | Deep ViT Features as Dense Visual Descriptors |
| SAPA | 10.62 | 4.85 | SAPA: Similarity-Aware Point Affiliation for Feature Upsampling |
| DIP | 10.57 | 5.16 | Deep Image Prior |
| CARAFE | 10.24 | 4.96 | CARAFE: Content-Aware ReAssembly of FEatures |
| FeatUp (JBU) | 9.83 | 5.24 | FeatUp: A Model-Agnostic Framework for Features at Any Resolution |
| FeatUp (Implicit) | 8.84 | 5.60 | FeatUp: A Model-Agnostic Framework for Features at Any Resolution |
0 of 6 row(s) selected.