Table 5 Performance comparison of different lightweight models on the self-built dataset.
From: SVTR-MG: an optical character recognition network for food packaging spray codes
Model | Accuracy (%) | FPS | Params (M) |
|---|---|---|---|
PP-OCRv3-Tiny | 84.7 | 260 | 5.1 |
ABINet-Tiny | 86.3 | 85 | 9.8 |
CRNN-Tiny | 87.2 | 120 | 7.3 |
ViTSTR-Tiny | 88.5 | 70 | 11.6 |
SVTR-T | 87.8 | 152 | 3.4 |
SVTR-MG | 93.2 | 142 | 142 |
