CLIP4STR / clip4str_huge_3e942729b1_log.txt
mzhaoshuai's picture
Upload CLIP4STR Pre-trained on DataComp-1B, LAION-2B, and DFN-5B
a70545b verified
Benchmark (Subset) set:
| Dataset | # samples | Accuracy | 1 - NED | Confidence | Label Length |
|:---------:|----------:|---------:|--------:|-----------:|-------------:|
| IIIT5k | 3000 | 99.53 | 99.86 | 97.82 | 5.09 |
| SVT | 647 | 99.07 | 99.78 | 97.26 | 5.86 |
| IC13_1015 | 1015 | 98.92 | 99.52 | 97.64 | 5.32 |
| IC15_1811 | 1811 | 91.72 | 97.41 | 94.43 | 5.37 |
| IC15_2077 | 2077 | 90.95 | 97.11 | 93.75 | 5.33 |
| SVTP | 645 | 97.98 | 99.46 | 96.43 | 5.86 |
| CUTE80 | 288 | 98.96 | 99.67 | 97.30 | 5.53 |
| HOST | 2416 | 82.57 | 95.34 | 87.73 | 5.38 |
| WOST | 2416 | 90.94 | 97.41 | 92.50 | 5.39 |
|-----------|-----------|----------|---------|------------|--------------|
| Combined | 14315 | 92.84 | 97.93 | 94.09 | 5.35 |
Benchmark set:
| Dataset | # samples | Accuracy | 1 - NED | Confidence | Label Length |
|:---------:|----------:|---------:|--------:|-----------:|-------------:|
| IIIT5k | 3000 | 99.53 | 99.86 | 97.82 | 5.09 |
| SVT | 647 | 99.07 | 99.78 | 97.26 | 5.86 |
| IC13_1015 | 1015 | 98.92 | 99.52 | 97.64 | 5.32 |
| IC15_1811 | 1811 | 91.72 | 97.41 | 94.43 | 5.37 |
| IC15_2077 | 2077 | 90.95 | 97.11 | 93.75 | 5.33 |
| SVTP | 645 | 97.98 | 99.46 | 96.43 | 5.86 |
| CUTE80 | 288 | 98.96 | 99.67 | 97.30 | 5.53 |
| HOST | 2416 | 82.57 | 95.34 | 87.73 | 5.38 |
| WOST | 2416 | 90.94 | 97.41 | 92.50 | 5.39 |
|-----------|-----------|----------|---------|------------|--------------|
| Combined | 14315 | 92.84 | 97.93 | 94.09 | 5.35 |
New set:
| Dataset | # samples | Accuracy | 1 - NED | Confidence | Label Length |
|:--------:|----------:|---------:|--------:|-----------:|-------------:|
| ArT | 35149 | 86.41 | 95.64 | 93.20 | 5.41 |
| COCOv1.4 | 9825 | 82.96 | 94.56 | 87.63 | 5.91 |
| Uber | 80551 | 91.71 | 96.59 | 92.23 | 5.36 |
|----------|-----------|----------|---------|------------|--------------|
| Combined | 125525 | 89.54 | 96.17 | 92.14 | 5.42 |
Time: Total 4051.573532819748s, Average 28.97292286055312ms. Total samples 139840.