bomtorazek

Sorry, I don't know the reason why the error occurs on your computer. It might be the problem with downloading the weights pre-trained on COCO, so you can download it...

> @bomtorazek Hi Did you reproduce the results ? Yes, as I mentioned above, I reproduced the results two years ago.

> > Sorry to bother you, but i have a problem about single channel image input. > > I did input one single channel image to main_train. But, An error...

> hello,thx for your source code > in the output image,what's the mean of G(z_opt).png and fake_sample.png > which one is the output from the net? If I'm correct, fake_sample...

scipy.stats.entropy uses the KL divergence if two distributions are given. "If qk is not None, then compute the Kullback-Leibler divergence" You can check the document below, https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.entropy.html

No, you can check figure 4 in the paper. The performance decreases as you reduce the batch size like the case for using the cross-entropy loss.

From test_caltech.py, change some lines. #if not os.path.exists(out_path): #os.makedirs(out_path) #files = sorted(os.listdir(w_path)) # get each files in w_path + sorting #for w_ind in range(51, 121): # get files from epoch...

Also I have changed the score from 0.01 to 0.5 following #38 boxes = bbox_process.parse_det_offset(Y, C, score=0.5,down=4) # originally 0.01

That's right. You can change it with the following: if(s==1), type='test'; skip=[]; else type='train'; skip=3; continue; end

res_path_test = 'data/cache/caltech2/test' image_data_gt, image_data_nogt, image_data_test = [], [], [] box_test_count = 0 boxes_test = [] boxes_test.append(box) boxes_test = np.array(boxes_test) box_test_count += len(boxes_test) if not os.path.exists(res_path_test): with open(res_path_test, 'wb') as...