ttt.txt

Web dataset radical length：33
Scene dataset radical length：39



Evaluation：

web SAN:
CUDA_VISIBLE_DEVICES=0,1,2 python main.py --config=configs/train_abinet.yaml --phase test --checkpoint=workdir/atop-ABINet-SAN-web-final/best-train-abinet.pth --test_root=data/web/web_val/ --model_eval=alignment --image_only

web Vision:
CUDA_VISIBLE_DEVICES=0,1 python main.py --config=configs/pretrain_vision_model.yaml --phase test --checkpoint=workdir/atop-Vision-SAN-web-final/best-pretrain-vision-model.pth --test_root=data/web/web_val/ --model_eval=vision --image_only

scene SAN:
CUDA_VISIBLE_DEVICES=0,1,2 python main.py --config=configs/train_abinet.yaml --phase test --checkpoint=workdir/atop-ABINet-SAN-scene-final/best-train-abinet.pth --test_root=data/scene/scene_val/ --model_eval=alignment --image_only

scene-Vision:
CUDA_VISIBLE_DEVICES=0,1 python main.py --config=configs/pretrain_vision_model.yaml --phase test --checkpoint=workdir/atop-Vision-SAN-scene-final/best-pretrain-vision-model.pth --test_root=data/scene/scene_val/ --model_eval=vision --image_only