Yogesh Kumar
Senior Researcher at Dolby Laboratories, Bengaluru
I work on video understanding: helping models find where, how, and when something happens in a video, given a query in text, an image, or another video. I recently joined the Advanced Technology Group at Dolby Laboratories, where I work on multimodal AI and long-form video understanding.
Before this, I did my M.Tech and PhD (defended in September 2026) at IIT Jodhpur in the Vision, Language, and Learning Group (VL2G), where I was fortunate to be advised by Dr. Anand Mishra. My thesis looked at grounding objects, relationships, and actions in videos with limited supervision. Earlier, I studied computer science at the University of Delhi, where I worked on text summarization with Prof. Vasudha Bhatnagar.
News
- Sep 2026Defended my PhD thesis, Where, How, and When: Query-Guided Grounding for Video Understanding, at IIT Jodhpur on 29 September. Thank you to Dr. Anand Mishra, my committee, and everyone at VL2G.
- Sep 2026Our paper on moment retrieval using video queries was accepted to Computer Vision and Image Understanding (CVIU).
- May 2026Joined the Advanced Technology Group at Dolby Laboratories, Bengaluru, as a Senior Researcher.
- Jan 2026Attended and volunteered at AAAI 2026 in Singapore, with travel support from ACM/IARCS.
- Jan 2026Started as a Senior Project Scientist at IIT Jodhpur, working on video-to-video retrieval.
- Dec 2025Presented our ICCV 2025 work on video moment retrieval at Vision India, ICVGIP 2025.
- Nov 2025Volunteered at EMNLP 2025 in Suzhou, China.
Show earlier news
- Nov 2025Recognized as an outstanding reviewer at BMVC 2025.
- Nov 2025Our work on few-shot video object detection was accepted to the AAAI 2026 main track.
- Jun 2025Our paper on video moment retrieval was accepted to ICCV 2025.
- Jun 2025Attended the Microsoft Research India Academic Summit 2025.
- May 2025Served as a reviewer for ICLR, NeurIPS, and BMVC 2025.
- Nov 2024Recognized as an outstanding reviewer at BMVC 2024.
- Sep 2024Our work on chapter-based video moment retrieval was accepted to ICVGIP 2024.
- Mar 2024Served as a reviewer for ECCV 2024.
- Jan 2024Attended Research Week with Google.
- Dec 2023Presented our CVPR 2023 work at Vision India, ICVGIP 2023.
- Dec 2023Our work on one-shot object localization in videos was accepted to AAAI 2024.
- Oct 2023Presented our work in the All India track at AIML-Sys 2023.
- Sep 2023Served as a reviewer for ICLR 2024.
- Apr 2023Our work on flowchart-to-code generation was accepted to ICDAR 2023.
- Mar 2023Served as a reviewer for NeurIPS 2023.
- Mar 2023Received travel support from Google, Microsoft, and ACM to attend CVPR 2023 in Vancouver, Canada.
- Feb 2023Our work on few-shot referring relationships was accepted to CVPR 2023.
- Jan 2023Presented my state-of-the-art seminar on interpreting video content using limited labels.
- Dec 2022Gave a tutorial on sequence models at the first Winter School on Responsible AI, IIT Jodhpur.
- Dec 2022Attended ICVGIP 2022 at IIT Gandhinagar.
- Sep 2022Presented my work at the Youth Conclave organized by INAE-SERB, IIT Jodhpur.
- Jun 2022Attended CVPR 2022 virtually.
- Jan 2022Selected to attend Research Week with Google in the computer vision track.
- Dec 2021Attended ICVGIP 2021 at IIT Jodhpur.
Selected publications
-
CVIU 2026
Moment Alignment Transformer for Moment Retrieval using Video Queries
Computer Vision and Image Understanding Preprint
-
AAAI 2026
-
ICCV 2025
-
EMNLP 2025
-
AAAI 2024
-
CVPR 2023
-
ICDAR 2023
A complete list is on Google Scholar.
Experience
-
May 2026 to present
Senior Researcher
Dolby Laboratories, Advanced Technology Group
Working on multimodal and long-form video understanding.
-
Jan 2026 to Apr 2026
Senior Project Scientist
IIT Jodhpur
Worked on video-to-video retrieval: given a short query clip, finding the matching segments across a collection of untrimmed videos, without going through text.
-
Jul 2024 to Jun 2025
Research collaboration with Microsoft Bing
IIT Jodhpur, supported by the Microsoft Academic Partnership Grant
Worked with colleagues at Microsoft Bing, Hyderabad, on searching long, untrimmed videos for actions and moments using text and video queries. Together we built MATR, a transformer that retrieves a moment using a video clip as the query, so the moment does not have to be described in words. The work appeared at ICCV 2025, ICVGIP 2024, and in CVIU, along with the SportsMoments and ChapVidMR benchmarks.
-
Jul 2023 to Jun 2024
Research collaboration with Accenture Labs
IIT Jodhpur, supported by an Accenture Gift Grant
Worked with colleagues at Accenture Labs, Bangalore, on localizing unseen objects and their interactions in videos from a handful of examples. This included QDETRv, a query-guided DETR for one-shot object localization, and a few-shot approach to grounding subject, predicate, and object relations across frames. The work appeared at CVPR 2023 and AAAI 2024.
-
Aug 2020 to Dec 2025
PhD Researcher and Teaching Assistant
IIT Jodhpur, VL2G Lab
Research in the Vision, Language, and Learning Group with Dr. Anand Mishra, alongside teaching assistance for seven courses.
Education
-
2020 to 2026
M.Tech and PhD in Artificial Intelligence
Indian Institute of Technology Jodhpur
- Thesis: Where, How, and When: Query-Guided Grounding for Video Understanding
- Adviser: Dr. Anand Mishra
- Supported by a UGC NET-JRF fellowship and academic grants from Microsoft and Accenture
-
2017 to 2019
Masters in Computer Science
University of Delhi
- Thesis: Extractive Text Summarization
- Adviser: Prof. Vasudha Bhatnagar
-
2014 to 2017
Bachelor (Hons.) in Computer Science
University of Delhi
Recognition and support
I am grateful to the people and organizations who have supported this work.
- Outstanding Reviewer, BMVC 2024 and 2025
- UGC NET-JRF in Computer Science, 2019 (All India Rank 17)
- GATE in Computer Science, 2020 (Rank 605)
- Research supported by the Microsoft Academic Partnership Grant (2024) and an Accenture Gift Grant (2023), both with Dr. Anand Mishra as Principal Investigator
- Travel support from ACM/IARCS to attend CVPR 2023 (Vancouver) and AAAI 2026 (Singapore), and from Google and Microsoft to attend CVPR 2023
- Attended Research Week with Google in 2022, 2024, and 2025
- Attended the Microsoft Research India Academic Summit 2025
Teaching
Teaching assistant, IIT Jodhpur
| Course | Term | Instructor |
|---|---|---|
| Pattern Recognition and Machine Learning | Jan to May 2025 | Dr. Anand Mishra |
| AI, Fundamentals of Machine Learning | Jul to Dec 2024 | Dr. Anand Mishra |
| Computer Vision | Jan to May 2024 | Dr. Pratik Majumder |
| Fundamentals of Machine Learning | Jul to Dec 2023 | Dr. Anand Mishra |
| Data Structure and Algorithms | May to Jul 2023 | Dr. Pratik Majumder |
| Computer Vision | Jan to May 2023 | Dr. Anand Mishra |
| Machine Learning | Jan to May 2022 | Dr. Richa Singh and Dr. Anand Mishra |
Tutorial
Seq2seq models, Winter School on Responsible AI, IIT Jodhpur, Dec 2022
Academic service
Reviewing
- Conferences
- CVPR 2026; ECCV 2024, 2026; NeurIPS 2023, 2025, 2026; ICLR 2024, 2025; BMVC 2024, 2025, 2026
- Journals
- TPAMI, TMM, TIST, TIP
Volunteering
- AAAI 2026, Singapore
- EMNLP 2025, Suzhou, China
- Winter School 2022 and 2024, IIT Jodhpur
- NCVPRIPG 2023, IIT Jodhpur
- International Colloquium on Ethics and Governance of Autonomous AI Systems, Feb 2019
Talks
- Video moment retrieval using video queries (our ICCV 2025 paper). Vision India, ICVGIP 2025.
- One-shot object localization in videos (our AAAI 2024 paper). Vision India, ICVGIP 2024.
- Visual relationship grounding in videos (our CVPR 2023 paper). Vision India at ICVGIP 2023, All India Track at AIML-Sys 2023, and Premier Papers Kaleidoscope at the ACM SIGKDD Workshop 2023.