算法名称 | 显卡 | 运行速度 | acc recall fmeasure | 思路 |
---|---|---|---|---|
Text-CNN | ICDAR-2013 0.93 0.73 0.82 | Text-CNN(attention text) + CE-MSERs detector | ||
CTPN | 0.14s/per | ICDAR-2013 0.93 0.83 0.88 | Side-refinement+blstm+vgg+Fine-scale proposals+bounding box | |
Text-boxes | 0.09s/per | ICDAR3end2end 0.91 0.89 0.84 | ssd+textboxes layers | |
SegLink | TITAN x*4 | 50ms/per | 85.3(f) | SSD+LINK+segment |
R2CNN | K80 | 0.39/per | ICDAR2015 63.60% 61.24% 62.40% | faster-rcnn+three Pooled size |
EAST | TITANX | 16.8(FPS) | ICDAR 2015 0.7833 0.8327 0.8027 | PVAnet+fcn+NMS |
算法名称 | 显卡 | 运行速度 | acc recall fmeasure | 思路 |
---|---|---|---|---|
CRNN | 0.16s/per | SVT 0.8 | CNN+BiLSTM+CTC | |
SEE | FSNS 0.78 | ResNet location CNN+BiLSTM+FC recognition |
||
Google Attention | TITAN X | FSNS 0.8379 | ssd+textboxes layers |