>>11124453>>11124470Ah thanks for the input. What I'm working is something similar to fingerprint case, actually it's finger vein patterns. So in my case I have over 100 people finger veins, my general idea is developing a system for authentication based on finger vein patterns. More conventional approaches is using pure computer vision techniques like local line binary pattern, repeatwd line tracking, phase correlation etc, but there are some papers that use deep learning algorithms like AlexNet, VGGnet, CNN. My friend is also working on same problem but with Capsule Network, so I was wondering if Residual Network will work too. I'm still not 100% convinced if this will work efficiently or whether I'll get satisfying results because just like you said, generally we need a pattern, or a class. What I'm thinking is this method is more suited for a case that has predefined classes, like classifying white blood cells.
