This sounds like something Karpathy did on selfies - http://karpathy.github.io/2015/10/25/selfie/ Good work to suggest filters to actually improve the photo! Not sure how you decide the filter selection though.
Squeeze nets have shown great promise, especially in reducing the model size so they can be deployed on embedded systems.
That plus the fact that the network was developed by a Chinese AI startup (and not by DeepMind or OpenAI) makes it very interesting.
why does it matter? chinese AI startup or not?
PS: SqueezeNets were introduced by DeepScale.ai Ref: https://github.com/DeepScale/SqueezeNet
I deeply apologize for the submission from my team. I know our earlier posts haven’t been of great quality but I believe this and our other blogs published recently have been received very well.
We will avoid posting anything on lobsters from now on.