PSENet.pytorch icon indicating copy to clipboard operation
PSENet.pytorch copied to clipboard

A pytorch re-implementation of PSENet: Shape Robust Text Detection with Progressive Scale Expansion Network

Results 26 PSENet.pytorch issues
Sort by recently updated
recently updated
newest added

请问有可以用的验证模型吗@author

恢复训练时,除了在config.py中更改三行 pretrained = True restart_training = False checkpoint = '/home/kpl/PSENet/outputs/13.pth' 如果有如下错误 RuntimeError: Error(s) in loading state_dict for PSENet:........................................ 可尝试在dataset/data_utils.py中 ![image](https://user-images.githubusercontent.com/37236583/157655265-54c8f315-a590-4f13-b348-f7a93120ab64.png) 将load_state_dict(state_dict) 改成 model.load_state_dict(state_dict, False)

Traceback (most recent call last): File "train.py", line 263, in main() File "train.py", line 217, in main recall, precision, f1 = eval(model, os.path.join(config.output_dir, 'output'), config.testroot, device) File "train.py", line 132,...

> > 看一下图片读进去没,如果你是在linux系统下,你最好进文件夹ls一下,看看是否有不是图片的文件 > > 谢谢,文件夹中多了一个自动生成的文件,删除解决了,但是在eval函数中(141-150行),boxes_list返回时空,导致报下面的错误: > > Traceback (most recent call last): > File "/mnt/data/ocr/psenet/PSENet.pytorch-master/cal_recall/rrc_evaluation_funcs.py", line 326, in main_evaluation > evalData = evaluate_method_fn(p['g'], p['s'], evalParams) > File "/mnt/data/ocr/psenet/PSENet.pytorch-master/cal_recall/script.py", line...

测试的时候,需要pse解码,我看代码中用的是c++的解码,pse.cpp,请问有没有大神有Python的版本?

`if random.random() < 0.5: im, text_polys = data_aug.horizontal_flip(im, text_polys) if random.random() < 0.5: im, text_polys = data_aug.random_rotate_img_bbox(im, text_polys, degrees)` 小于0.5要进行两次操作?

大神好! 我看[这里](https://github.com/WenmuZhou/PSENet.pytorch/blob/e0ee2ed852259d0e7d036b071119d53a11a0b84d/train.py#L131)只是cv2.imread 了进来BGR的图片. 但是[data generator](https://github.com/WenmuZhou/PSENet.pytorch/blob/e0ee2ed852259d0e7d036b071119d53a11a0b84d/dataset/data_utils.py#L106) 里是有BGR2RGB的. 我实际测试下来 加了 BGR2RGB的F1 会高1%. 请看看在eval model 的时候是不是需要加BGR2RGB

How to train ctw1500 dataset? Thank you very much.

hey thanks for great work. i am training this model on icdar 2015 dataset. i saw you save checkpoint when loss will be in 0.3 to 0.4 and epoch %...