I am working with pytorch, and trying to implement a semantic segmentation method.
I want to visualize a feature map whose size is [10,192,128,128], 10 is the batch size, 192 is the number of channels and (128,128) being spatial dimension.
Is there anyway to visualise this feature map using tensorboard in pytorch.??
I tried using add_image(), but I think it works only with 3 channels.
You could visualize each channel for all samples as a grayscale image (which would result in 1920 images).
torchvision.utils.make_grid might be useful to create a smaller grid of these images.