sketchRetrieval
Freehand sketches are a simple but powerful tool for communication. They contain rich information to specify shapes, and thus retrieving 3D shapes from 2D sketches has received considerable attention in the field of computer graphics and computer vision. In this project, we present a system for cross-domain similarity search that helps us with sketch-based 3D shape retrieval. Instead of using hand crafted features for searching, we propose our DeepSketch neural network that is built on Siamese network to learn features that are basis for later similarity search using K-nearest Neighbor (KNN). We further did analysis on how individual strokes of a sketch image affect retrieval results, and also visualized features learned from our DeepSketch network.