Shreyash Pandey

I am a graduate student in the department of Electrical Engineering at Stanford University. I am primarily interested in Deep Learning and its application to Computer Vision and Natural Language Processing.

I am currently working on weak supervision within the Hazy Research group under the supervision of Professor Christopher Re.

Before coming here, I worked for a year at Samsung Research Institute, Bangalore in their VizInsight Computer Vision Research division. My work revolved around image classification and object detection using deep learning technologies. Prior to that, I did my undergraduation from Indian Institute of Technology Kanpur with a major in Electrical Engineering.

Email  /  LinkedIn  /  Resume

Notable Projects

Robust Deep RL for Autonomous driving

Implemented Deep Deterministic Policy Gradient to autonomously drive a race car based on TORCS simulator.

Checked it's robustness to additive white Gaussian noise, and suggested ways of making it more robust.



Deep Reinforcement Learning for Atari Games

Implemented DQN from scratch for Atari games such as Space Invaders and Q*bert.

Compared the performance with Double DQN and dueling architectures.



Instagram Hashtag Prediction, With and Without Data

Built a data collection and cleaning pipeline for fully supervised image classification engine for Instagram Hashtag Prediction.

Implemented the state-of-the-art Zero Shot Learning approach to transfer a pre-trained ResNet to predict Instagram Hashtags with comparable Top-5 accuracy.



A survey of Zero Shot Learning
Supervisor: Dr. Piyush Rai, Indian Institute of Technology Kanpur

Studied different methods of performing Zero Shot Learning(ZSL) - prediction of a label that has been not seen during the training procedure.

Implemented two contemporary papers from this area which required learning a common semantic space for embedding images and labels, to perform ZSL task. Focused on dictionary learning as a way to resolve the PDS issue and found that CNN based features drastically improve the classification accuracy.

report | poster


Visualization of high dimensional data
Supervisor: Dr. Ketan Rajawat, Indian Institute of Technology Kanpur

Explored applications of convex optimization for dimensionality reduction, especially over non linear manifolds.

Compared performance based on visualizations, computational complexities, and error rates obtained in classification tasks. Selected as the best project in the course comprising of over 80 students.



Samsung Research and Development Institute, Bengaluru (SRIB)
Supervisor: Vishwanath Gopalakrishnan, Principal Engineer, SRIB

Developed a photo search mobile application based on image classification by using memory efficient CNN architectures.

Implemented a depth prediction module for 2D images by formulating it as a dense-label regression problem.

inspired from this website