Sateesh Kumar

I completed my Masters at University of California San Diego (UCSD) in Summer 2023. I researched on video understanding and robotics and was advised by Prof. Xiaolong Wang.

After finishing my masters, I joined TikTok as a Researcher from July 2023. I work in the TikTok AI Generation team.

Previously, I worked at Retrocausal for 2 years as a Research Engineer (Computer Vision) under the supervision of Dr. Quoc Huy Tran and Dr. Zeeshan Zia.

I completed my bachelor's in Computer Science from FAST NUCES Karachi, where I worked on Class Imbalance under the guidance of Prof. Tahir Syed.

Email  /  CV  /  Google Scholar  /  Twitter

profile photo
Publications and Preprints

Papers are in reverse chronological order. '*' denotes equal contribution.

The Devil is in the Details: A Deep Dive into the Rabbit Hole of Data Filtering
Haichao Yu, Yu Tian, Sateesh Kumar, Linjie Yang, Heng Wang
International Conference on Computer Vision (ICCV) DataComp Workshop , 2023
(Ranked 1st in DataComp challenge)

We introduce a three-stage filtering strategy for enhancing model performance. It focuses on single-modality filtering, cross-modality filtering, and data distribution alignment. The proposed approach significantly surpasses previous methods on the DataComp benchmark.

Graph Inverse Reinforcement Learning from Diverse Videos
Sateesh Kumar, Jonathan Zamora*, Nicklas Hansen*, Rishabh Jangir, Xiaolong Wang
Conference on Robot Learning (CoRL) , 2022 (Oral)
project page / arXiv

GraphIRL is a self-supervised method for learning a task reward solely from videos. We build an object-centric graph abstraction from video demonstrations and then learn an embedding space that captures task progression in a self-supervised manner by exploiting the temporal cue in the videos.

Unsupervised Action Segmentation by Joint Representation Learning and Online Clustering
Sateesh Kumar*, Sanjay Haresh*, Awais Ahmed, Andrey Konin , M. Zeeshan Zia, Quoc-Huy Tran
CVPR, 2022
project page / arXiv

We propose temporal optimal transport for jointly learning representations and performing online clustering in an unsupervised manner. The approach learns prototype vectors via backpropogation. The prototype vectors are initialized at random and act as cluster centroids.

Learning by Aligning Video in Time
Sateesh Kumar*, Sanjay Haresh*, Huseyin Coskun, Shahram N. Syed, Andrey Konin , M. Zeeshan Zia, Quoc-Huy Tran
CVPR, 2021
project page / arXiv

We propose alignment as pre-text task for self-supervised video representation learning. The proposed approach leverages differentiable dynamic time warping for learning global alignment across pairs of videos.

Towards Anomaly Detection in Dashcam Videos
Sateesh Kumar*, Sanjay Haresh*, M. Zeeshan Zia Quoc-Huy Tran
IV, 2020
talk / arXiv

We collect a video dataset of road-based anomalies. We propose an object-object interaction reasoning approach for detecting anomalies without additional supervision. We experiment with reconstruction based and one-class classification based approaches.

Website layout is from Jon Barron