Yash Kumar Sahu

I'm a research assistant at I3D Lab in Indian Institute of Science (IISc) advised by Prof. Pradipta Biswas. At I3D, I’ve worked on diffusion-based image editing, 3D reconstruction, and object affordance prediction. I did my Bachelors and Masters in Computer Science at Indian Institute of Information Technology, Design and Manufacturing (IIITDM), Kancheepuram. Previously, I was a research student at GCDSL-MRL Lab, working on 3D path planning for drones with Prof. Debasish Ghose.

Email / CV / Linkedin / Scholar / Github

News

[05/2025]: Check out our I3D-Tools-Dataset at Hugging Face 🤗.
[01/2025]: 1 paper, "Blind Tactile Exploration for Surface Reconstruction" has been accepted at ICRA 2025.
[12/2024]: 1 paper, "Diffuse Your Data Blues: Augmenting Low-Resource Datasets via User-Assisted Diffusion" has been accepted at IUI 2025. Featured in a press release at Times of India.

Research

I'm interested in computer vision, deep learning, generative AI, and image processing. Most of my research is about inferring the physical world (shape, position, alignment, semantics etc.) from images.

	Synthetic Tools Dataset via Diffusion Models Yash Kumar Sahu, Yashaswi Sinha, Himanshu Vishwakarma, Arushi Khokhar, Pradipta Biswas NeurIPS*, 2025 (Under Review) DATASET A 37,000-image dataset of hand tools composited on diverse generated backgrounds via img-to-img diffusion, with detailed scene captions, annotations, and segmentation maps.
	Blind Tactile Exploration for Surface Reconstruction Yashaswi Sinha , Soumojit Bhattacharya , Yash Kumar Sahu, Pradipta Biswas ICRA, 2025 VIDEO Exploring and 3D reconstructing complex shapes using only tactile feedback from a single touch, without relying on vision.
	Diffuse Your Data Blues: Augmenting Low-Resource Datasets via User-Assisted Diffusion Yashaswi Sinha, Yash Kumar Sahu , Shravan Shanmugam , Abhishek Mukhopadhyay, Pradipta Biswas ACM IUI, 2025 VIDEO Customize backgrounds and foregrounds in diffusion-based image generation, supporting tasks like image data augmentation to enhance object detection accuracy.
	A Comparative Study on Image Translation GAN Models to Improve Object Detection Accuracy on Low-Resource Domains Yash Kumar Sahu , Abhishek Mukhopadhyay, Gyanig Kumar, Pradipta Biswas ICTTVS*, 2024 PPT Improving object detection in scenarios with limited number of images by augmenting that dataset with GAN generated synthetic images.
	Vision-Based Object Sorting in Dynamic Environments using YOLO for RoboCup ARM Challenge 2023 Yash Kumar Sahu , Radhika Mittal, Deep Paresh Patel, Chayan Maiti, Sreekumar Muthuswamy CICT*, 2023 PPT Performing autonomous robotic manipulation for sorting objects in constantly changing environments.

Robotics Challenges

	ICRA 2024 - RoboMaster University Sim2Real (RMUS) Challenge Won 3rd Prize globally at the finals held in Japan. Competing among 30+ teams (1st ever Indian team to reach finals). Organised by the Tsinghua University.
	RoboCup 2023 - Autonomous Robot Manipulation (ARM) Challenge Ranked 4th globally and 3rd in classification accuracy among 10+ countries in the finals held in France. Sponsored by MathWorks.
	European Rover Challenge (ERC) 2022 Ranked 6th globally at the remote edition world finals featuring 50+ teams from 10+ countries. Organised by the European Space Agency (ESA).
	RoboWars 2023 - IIITDM Vashisht Technical Festival Awarded 2nd prize inter-university in the finals competing in a physical battle against 12 robots. Organised by the Robotics Club at IIITDM.