In-Image Machine Translation Evaluation on Vectra-Bench zh→ja 1.0
0.677Pearson Correlation (r)Vectra
Evaluation results
| Method | Links | ||
|---|---|---|---|
| VectraProtocol=No-Reference (Direct Assessment)2026.01 | 0.677 | 0.599 | |
| FIDProtocol=Reference-based (Top-Ranked Candidate as Reference)2026.01 | 0.559 | 0.465 | |
| HCIIT-JudgeProtocol=No-Reference (Direct Assessment)2026.01 | 0.538 | 0.46 | |
| SSIMProtocol=Reference-based (Top-Ranked Candidate as Reference)2026.01 | 0.537 | 0.438 | |
| PSNRProtocol=Reference-based (Top-Ranked Candidate as Reference)2026.01 | 0.525 | 0.431 | |
| SAMScoreProtocol=No-Reference (Direct Assessment)2026.01 | 0.401 | 0.306 | |
| LPIPSProtocol=Reference-based (Top-Ranked Candidate as Reference)2026.01 | 0.331 | 0.268 | |
| E-Commerce VQAProtocol=No-Reference (Direct Assessment)2026.01 | -0.08 | -0.068 |