Deep Learning Technique for Image Colorization.

Image colorization has its importance in research, such as thief identification from sketch images, automatic colorization of old pictures, etc. Image colorization is considered a troublesome issue and regularly requires manual acclimation to accomplish the desired quality. Colorization should be possible in either an interactive manner or an automatic manner. Motivated by the ongoing achievement in deep learning procedures that give astounding displaying of enormous scope information, creators present a Deep Convolutional Neural Network approach for automatic image colorization. A grayscale image is passed as an input into the network, and based on the local and global features, an equivalent colored image is generated as an output by the trained network. The model has trained with the Places dataset, which has approximately 100000 images with 100 different classes.