Human Judgment Classification 벤치마크
Human Judgment Classification on Pascal-50S
Mean Accuracy
- 2021-04-18 — RefCLIP-S: Mean Accuracy 83.1
- 2022-05-25 — MID: Mean Accuracy 85.2
| Rank | Model | Mean Accuracy | Paper | Code | Year |
|---|---|---|---|---|---|
| 1 | MID | 85.2 | Mutual Information Divergence: A Unified Metric for Multimodal Generative Models | naver-ai/mid.metric | 2022 |
| 2 | RefCLIP-S | 83.1 | CLIPScore: A Reference-free Evaluation Metric for Image Captioning | jmhessel/clipscore · showlab/loveu-tgve-2023 · jmhessel/pycocoevalcap | 2021 |
| 3 | CLIP-S | 80.7 | CLIPScore: A Reference-free Evaluation Metric for Image Captioning | jmhessel/clipscore · showlab/loveu-tgve-2023 · jmhessel/pycocoevalcap | 2021 |