siyuhuang / crowdcount-stackpool Goto Github PK

I have read your papers. But I don't understand what the Figure2 tries to explain and what the colors(gray, blue, and orange) represent. Can you give me some ideas about that? Thank you!

1. Your paper says that we randomly crop 9 patches on each training image, where all the patches are half the size of the original image. Actually after running create_training_set_shtech.m, the patches' height and width are about a quarter of the original images.

2. crowdcount-stackpool/data_preparation/create_training_set_shtech.m

   Lines 49 to 61 in 403be5a

   	wn2 = w/8; hn2 = h/8;
   	wn2 =8 * floor(wn2/8);
   	hn2 =8 * floor(hn2/8);
   	annPoints = image_info{1}.location;
   	if( w <= 2*wn2 )
   	im = imresize(im,[ h,2*wn2+1]);
   	annPoints(:,1) = annPoints(:,1)*2*wn2/w;
   	end
   	if( h <= 2*hn2)
   	im = imresize(im,[2*hn2+1,w]);
   	annPoints(:,2) = annPoints(:,2)*2*hn2/h;
   	end

   
   What does the code try to do? The condition w <= 2*wn2 is equivalent to w<=16*floor(w/64). But only when w=0 or w<=-16, the condition meets.

in your density group in figure 5, which number do you select to classifiy density group

When I test my trained model on the test data of ShanghaiTechA, there are 92 images, each of which takes more than 1 s and meanwhile there are 61 images, each of which takes less than 0.1 s. My GPU is GTX 1070.
My test method is

torch.cuda.synchronize()
start = time.time()
density_map = net(im_data)
density_map = density_map.data.cpu().numpy()
torch.cuda.synchronize()
end =  time.time()

Why does it happen? How can I solve it? Thank you!

when I run train.py it misses the train_path, I'm a freshman in this area so can you help me to train zhe dataset?

Hello, is the transformation density map using a geometric adaptive kernel?

hi~
Can you share how did you deal with the number difference between the gt density map before and after downsample?
Specifically, when I trained the model, I found the number of people (groundtruth) calculated by the ground truth density map before and after downsampling is different. So, if I train with the downsampled gt map, the ground truth number in the final MSE is the sum of downsampled map or the original gt number?
Thank you.