Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
The output is passed through two convolution layers and then a sigmoid activation is applied to predict the depth of each pixel.