| here's my side project I just finished: http://www.eggie5.com/126-semantic-image-search-video Semantic Image Search: takes high dimensional images vectors from a Deep Convolutional Neural Network and searches for similar images using Locality Sensitivity Hashing. Search is O(1) + O(n/B) where the last term is a small fraction, B, of the complete search space n. The goal was an E2E machine learning project: modeling, training, validation, deployment and serving. Using VGG-mased CNN w/ tensorflow, redis for database, Flask for web frontend , tensorflow serving and docker+kubernetes for orchestration. It all works on my local machine but when I went to deploy it to my prod website the tensorflow serving part broke down and I haven't had time to fix it yet :( It was just supposed to be a one weekend project after I got the inspiration from the AlexNet paper, but it took over a month. Most of the time was fighting k8s. So another data science project doesn't make it to production... |