Colouring black and white images using deep learning
I tried the above two methods on wider face dataset both shows good results.
Using resuidal blocks reduces time in training and the trainable parameters in are also very much less model that cancatenates the output of vgg layer.