ViSiL: Fine-grained Spatio-Temporal Video Similarity Learning - View it on GitHub