Image Segmentation

  • Using convolutional layers to make predictions for all pixels at once - This method is very expensive to do on original image resolution.

  • Convolutions at original image resolution

  • To efficiently do segmentation, design the network as a bunch of convolutional layers, with downsampling and upsampling inside the network!

  • Downsampling and upsampling

  • Using downsampling, upsampling and using cross-entropy loss for the pixels, we can do image segmentation

Reference:-