2
$\begingroup$

I'm looking for neural network architecture that excel in counting objects. For example, CNN that can output the number of balls (or any other object) in a given image. I already found articles about crowd counting, I'm looking for articles about different types of objects.

Thank

Improve this question
$\endgroup$
1
$\begingroup$

If you want to count the number of objects using a neural network, you can use pretrained YOLO with the bottom prediction layer removed, and feed the features to a classification feed forward layer of let's say 1000 class representing 0-999 objects in the image. You can then train it and propagate the gradients through it. For example, in the pytorch code for YOLO,(source:https://github.com/eriklindernoren/PyTorch-YOLOv3) You can add a nn.Linear and use cross entropy loss to classify the number of images. You can also change the architecture completely. Maybe you can try adding layers to reset or other classifying network to count the number of objects. Hope this can help you and have a nice day!

Improve this answer
$\endgroup$
1
$\begingroup$

You can run YOLO, then count the number of occurrences.

Improve this answer
$\endgroup$
  • $\begingroup$ Thanks, but I want to use this CNN as part of a larger architecture that contains more NN. So in order to train the larger architecture I want the count network to be differentiable, so I could backpropagate. Thus, I can't just "count" the number of occurrences. $\endgroup$ – ron653 Mar 11 '19 at 8:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.