Large Scale Image Retrieval with Locality Sensitive Hashing

With the ever growing size of image databases today and the increasing demand for search and recommendations from within those databases, it is becoming more and more critical to find efficient image search techniques. This paper proposes a fast approach for extracting the most similar image to a given query image from a database such that the extracted images are most similar semantically. This involves reducing the dimensions of the images by encoding them uniquely to lower dimensional vector using CNN and then applying locality sensitive hashing along with euclidean distance similarity to find the most similar images.