Udit Singh Parihar

Udit Singh Parihar

Deep Learning Researcher

Robotics Research Center, IIITH

Biography

Hey! I am currently working as a Research Assistant under Prof. Madhava Krishna in Robotics Research Center. My current research statement includes making robust Simultaneous Localization and Mapping (SLAM) algorithms in a featureless environment and improving correspondence matching in high illumination and viewpoint changes. We tackle these problems by incorporating classical computer vision in deep learning pipelines.

Download my resumé.

Interests

  • SLAM / State Estimation
  • Differentiable Computer Vision
  • Optimization
  • Robotics

Education

  • MS by Research in CSE with specialization in Robotics, 2021

    International Institute of Information Technology, Hyderabad

  • B.Tech in Mechanical, 2018

    Indian Institute of Technology Jodhpur

Publications & Preprint

Quickly discover relevant content by filtering publications.

La-MAML: Look-Ahead Meta-Learning for Continual Learning

In this work we develop a gradient-based meta-learning algorithm for efficient, online continual learning, that is robust and scalable to real-world visual benchmarks.
PDF Code Project Video Oral
La-MAML: Look-Ahead Meta-Learning for Continual Learning

Probabilistic Object Detection: Strenghts, Weaknesses, and Opportunities

This work surveys recent work in the very nascent field of probabilistic detection and pesents insights and promising avenues for future research in this area.
PDF Project Slides Source Document
Probabilistic Object Detection: Strenghts, Weaknesses, and Opportunities

Unifying Variational Inference and PAC-Bayes for Supervised Learning that Scales

This work proposes to train models with imitation learning in such that they come with a PAC-bayes bound as a performance guarantee for the model. It connects the expressions for the PAC bayes bound and the ELBO of a stochastic predictive policy learnt through likelihood maximisation. The model is then trained by minimising the pac-bayes bound as the objective that also doubles as an error bound.
PDF Code

Viewpoint Invariant Junction Recognition using Deep Network Ensembles

This work introduces and formalises the problem of recognising intersections from drastically different viewpoints to enable place recognition for SLAM using a siamese convolutional recurrent architecture trained to classify pairs of short video streams.
PDF Video
Viewpoint Invariant Junction Recognition using Deep Network Ensembles

Geometric Consistency for Self-Supervised End-to-End Visual Odometry

CTCNet proposed using the compositional property of transformations to self-supervise learning of visual odometry from images.
PDF Code Project Poster
Geometric Consistency for Self-Supervised End-to-End Visual Odometry

Projects

*
All Deep Learning Robotics Computer Vision Other

Autonomous Driving using IRL

Learning from Demonstrations with Inverse Reinforcement in CARLA

Training diverse ensembles for OOD object detection

Exploring Stein Variational Inference for ensemble training

Stein methods for Robot Navigation

Poster ICML workshop on stein’s method

Contact

  • IIITH, Hyderabad, Gachibowli 500032