for classification of images how to decide which architecture would be better resnet or vgg or alexnet
If you are talking about transfer learning i.e. using pre-trained CNN architectures for image classification then this might help: https://cs231n.github.io/transfer-learning/ .
ResNet would most likely perform best and is the one I’d choose out of those alternatives
okay it is the best but what is the need to use that every time.if some simple network can give the same accuracy in less time.
Well try it and see but there’s no way to know apriori and ResNet would most likely (like everytime), outperform especially compared to something like AlexNet. There’s no need to use it everytime, but people want good performance and that’s why they use those architectures. You can try it out and make your own opinion.