search

ylf-li / ContourGAN

Tensorflow for ContourGAN: Image Contour Detection with Generative Adversarial Encoder-Decoder Networks
73 stars 16 forks source link

IOError: /opt/data/HED-BSDS/train_pair.txt not found. #2

Open JeremyJian opened 6 years ago

JeremyJian commented 6 years ago

Hi, Thanks for your sharing. When I run this project. I ran into "IOError: /opt/data/HED-BSDS/train_pair.txt not found." Where can I get the train_pair.txt file?

GeneralJing commented 6 years ago

same question as above,

shreelock commented 6 years ago

I think that is the file available in the HED-BSDS dataset(link), by the name of train_pair.lst. I just renamed it and used the same.

GeneralJing commented 6 years ago

thank you. I met another error, as follows: gts_raw=[cv2.imread(x[1],0)/255.0 for x in imgbatch];gts=np.array(gts_raw) TypeError: unsupported operand type(s) for /: 'NoneType' and 'float'. I checked the jpg and png image, it seemed no problem. Do you know this kind of questions?

shreelock commented 6 years ago

The image is not being read properly, you would have to correct the image path. imgbatch is has relative path of images. You'd have to correct it to take absolute paths of all the images. Probably need to append dataset location.

GeneralJing commented 6 years ago

Thank you.That question was settled,but I ran into another pit. gts_raw=[cv2.imread(x[1],0)/255.0 for x in imgbatch];gts=np.array(gts_raw) gts[gts>0.5]=1.0 ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all(). I checked a lot of information,however, it has not been resolved yet. I usually use C++, not familiar with python. can you give me some advice?

GeneralJing commented 6 years ago

is this problem caused by versions of python2 or 3? anyone knows?

shreelock commented 6 years ago

which line is the error coming from? is it gts[gts>0.5]=1.0 ? The error is raised by numpy. This step is just thresholding gts. If the error is caused by this line, then you can iterate through all the elements in gts, and set their value to 1.0 if its more than 0.5. See this

jain-harshali commented 6 years ago

what is params in demoGAN.py in line 87 (model[params][:26]) i met with this error 'key error params is not a file in the archive'.

shreelock commented 6 years ago

Did you download the trained model.npz, as was mentioned in the README.md? Looks like problem might be coming from there. Otherwise, you would probably have to train the model.