RCF-pytorch evaluate.py?

evaluate.py?

Open rcorin opened this issue 5 years ago • 15 comments

hello, i can't find the evaluate.py file to test the project, do you know where it is?

also, do you have a pretrained model somewhere?

thanks

Oct 29 '18 13:10 rcorin

you can google Structured Edge Detection Toolbox. we use it to evaluate the result both in CVPR version of RCF (caffe) and in the pytorch version of RCF. I will consider writing a CUDA version of evaluating code in the later time. Sorry for missing the pretrained model, because I am working for the deadline. I will update it after the deadline.

Oct 29 '18 15:10 meteorshowers

@meteorshowers could you upload your pretrained model?, How long did you training in BSDS500 dataset?

Jan 15 '19 13:01 herleeyandi

yun-liu has a pre-trained model you may want to take a look at: http://mftp.mmcheng.net/liuyun/rcf/model/5stage-vgg.caffemodel It is trained for 12 epochs, so make sure to change the "start" epoch in the train_RCf.py if you choose to use it

Feb 27 '19 04:02 wadewhitten

could you please open sourse evaluate.py?thank you

Jun 05 '19 02:06 ChengYingo

@meteorshowers hello, thank you very much for offering RCF-pytorch at first. At the same time, I have a question that I got evaluation index ODS=0.788 after using NMS insted of 0.808, where the pretrained model I used is fast-rcnn-vgg16-pascal07-dagnn.mat . I want to know what the reason is, and whether you can supply the pretrained model you utilize. More guidance from you will be highly appreciated.