Image Classification on QuickDraw (test)
87.5AccuracyProtoVAE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ProtoVAEprototypes per class=102022.10 | 87.5 | — | — | |
| Black-box encoderbackbone=same as ProtoVAE, type=classical feed-forward CNN2022.10 | 86.7 | — | — | |
| GCN-HViT-12026.04 | 86.55 | — | — | |
| GCN-HViT-2position embeddings=bidirectional adjacency matrix2026.04 | 86.35 | — | — | |
| HViT2026.04 | 86.12 | — | — | |
| ViT-16patch size=162026.04 | 84.87 | — | — | |
| ViT-4patch size=42026.04 | 83.65 | — | — | |
| FLINTbackbone=same as ProtoVAE2022.10 | 82.6 | — | — | |
| SENNbackbone=same as ProtoVAE2022.10 | 79.3 | — | — | |
| ProtoPNetbackbone=same as ProtoVAE2022.10 | 58.7 | — | — |