Zenglin Shi
Zenglin Shi
I'm not really understand your question. Basically, to train your model, you need first generate your density map ground truth. Then you need design your network, in which the output...
Yes, as you said, the last layer of convolutional output can be directly drawn into an density maps image. It would be better to understand if you take some time...
Hi, the NCL loss layer has been added in include and src folds. Thanks for your interest.
Hi, for my experience, some of your density map ground truth are generated incorrectly, their values are Nan, please note some annotations of shanghai partA are not correct, such as...
Sorry, no code, you can do that.
Great
You can refer to readme, which has some information about how to prepare your data. The codes for preparing crowd counting dataset can be found in ¨examples/crowd/shanghaiA/predata¨.
Hi, I just run the code, and I didn't have your error. x_shape = tf.shape(x), which dynamically gets the tensor shape during running, refer to https://pgaleone.eu/tensorflow/2018/07/28/understanding-tensorflow-tensors-shape-static-dynamic/ Could you provide your...
Hi, Please refer to https://github.com/gjy3035/Awesome-Crowd-Counting
Actually, MAE should be abs(sum(a)-sum(b)), in our paper, we use caffe_cpu_asum to compute a abs(sum(abs(a))-sum(abs(b))), and a little different in terms of the equation. However, we use a relu(a) before...