Natural language image descriptor

Generating descriptions for visual data (images and video) automatically has been a complicated task in the field of Computer Vision and Artificial Intelligence. This paper discusses the working of and improvements on an algorithm called Neural Image Captioner (NIC) by Oriol Vinyals and his team, which uses a deep convolutional and recurrent architecture to generate natural language sentences to describe the visual data input. We look at the possibility of making this algorithm train faster without allowing it to lose accuracy via the usage of techniques like Stochastic Gradient Descent and also employ an algorithm to find the perfect depth of the convolutional part of the network for different datasets. A drop of 33% was observed in the number of iterations required to get the algorithm to its original proficiency as claimed by Oriol et al.