Projects and resources for ramping up on computer vision
http://vision.stanford.edu/teaching/cs231b_spring1415/slides/alexnet_tugce_kyunghee.pdf
Train a binary classifier to distinguish between images of shoes you like and shoes you don't like. (Warning: this will require you to find many pictures of shoes.)
Train a digit classifier on the MNIST database (publically available online.)
https://research.googleblog.com/2014/11/a-picture-is-worth-thousand-coherent.html
Train a network that can distinguish between floors and walls in synthetic data.
Recreate deep dream: https://www.youtube.com/watch?v=MrBzgvUNr4w&list=PL2-dafEMk2A7EEME489DsI468AB0wQsMV&index=8 (this youtube series is useful.)
Furniture recognition: https://arxiv.org/pdf/1603.08637.pdf Face Recognition: https://medium.com/@ageitgey/machine-learning-is-fun-part-4-modern-face-recognition-with-deep-learning-c3cffc121d78
https://cmusatyalab.github.io/openface/ - train to recognize your family's faces
ORB-SLAM. One of the best algorithms out there. http://webdiis.unizar.es/~raulmur/orbslam/
This is tricky - SLAM is usually a large compilation of algorithms. Get one of the following parts to work:
Optical flow
Feature matching between frames
Get orb-slam running with a webcam. See if you can make it better.
https://grail.cs.washington.edu/rome/
Try to implement your favorite paper