Colorization of grayscale images is a multi-modal problem which has numerous use cases in the fields of art, science and medicine. This project presents an ensemble approach of deep convolutional neural network and ResNet high level feature extractor to colorize black-and-white images of fruits, landscapes and animals without human assistance. The final output image is compared with the original colored image and scored according to the mean squared error formula to determine the efficacy of our approach. Our model achieves close to 10 times lower MSE compared to a baseline model consisting of an 8-layer CNN with no high level feature extraction.