Scene Text Image Captioning on ViTextCaps (test)
0.413BLEU-1M2
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| M2Copy mechanism=false2026.04 | 0.413 | 0 | 0.225 | 0.151 | 0.026 | |
| BUTDCopy mechanism=false2026.04 | 0.409 | 0 | 0.253 | 0.196 | 0.048 | |
| HumanType=Reference Point2026.04 | 0.3 | 0.118 | 0.218 | 0.272 | 0.808 | |
| GPT-4oType=Reference Point, Copy mechanism=false2026.04 | 0.283 | 0.106 | 0.282 | 0.255 | 0.14 | |
| PhonoSTFGCopy mechanism=true2026.04 | 0.251 | 0.098 | 0.174 | 0.227 | 0.646 | |
| M4CCopy mechanism=true2026.04 | 0.244 | 0.087 | 0.172 | 0.221 | 0.594 | |
| HSTFGCopy mechanism=true2026.04 | 0.222 | 0.085 | 0.164 | 0.219 | 0.64 | |
| AOACopy mechanism=false2026.04 | 0 | 0 | 0.006 | 0.031 | 0.007 |