pan_pp.pytorch
pan_pp.pytorch copied to clipboard
Official implementations of PSENet, PAN and PAN++.
Thank you for your great work, I have three types text on my documents so I need to detect the type of each text line. Do you train your model...
 求助大家,这是什么原因呢
I train the model on ctw1500,MSRA-TD500,Total-text but i'm trying to do visualization can anyone help me with this please
运行命令是这个python ./eval/ic15_end2end_rec/script.py -g=./eval/ic15_end2end_rec/gt.zip -s=./outputs/submit_ic15.zip -o=./output -p={\"IOU_CONSTRAINT\":0.8}然后出现b'Line in sample not valid. Sample: 324 Line: 513,501,522,467,599,488,590,521 Error: Format incorrect. Should be: x1,y1,x2,y2,x3,y3,x4,y4,transcription'这个上传文件格式不对得问题,检查了GT里面得标注,没有找到这个文件
请问怎么解决啊谢谢
Hello, author. When I execute test according to the command you gave me, I have the following problem  How can I solve this problem?
想问下大佬,有知道 这个怎么绘制PR曲线的吗,只能训练完判断最终模型的指标吗?发现每个模型都不一样,有时候前面的比后面的准确值还高
您好,想请教一下,这个PAN和DBnet相比,哪个对中文效果检测更好呢?或者是这个论文对比DB有哪些优点呢?
复现效果问题
您好,我使用您的PAN++模型单独在IC15数据集上进行端到端训练,没有采用其他的数据集进行联合训练,所得的准确率与您的相近,也在83%左右,但召回率差得太多,仅有10%左右,导致最终的F-score不佳,请问大概是什么原因呢?