Deepak Sridhar

San Diego, California

I am a fifth year PhD candidate at UC San Diego specializing in Computer Vision, advised by Prof. Nuno Vasconcelos. My research focuses on multimodal foundation models, with particular interests in multimodal generation, editing, and understanding, as well as efficient personalization methods. My PhD thesis focuses on “Efficient Adaptation of Foundation Models” for fine-grained tasks. My work centers on the question: how do we get more capability out of multimodal foundation models with less training and compute?

During my PhD, I have developed an efficient diffusion model framework (project page, NeurIPS’24), explored efficient prompting techniques (project page ECCV’24, project page NeurIPS’26) and worked on efficient video reasoning (project page, CVPRF’26). Prior to my PhD, I have also worked on fundamental problems in computer vision such as efficient image classification, object detection, action recognition and localization.

Outside of Research, I love to go on hiking and exploring the nature. I also love to participate in outdoor adventures.

Canadian Citizen, TN-eligible

news

Sep 24, 2026	Our Paper “Diffusion Meta-Prompting and Steering for Generalizable Foundation Model Adaptation” is accepted to NeurIPS’26! This paper was an outcome of Qualcomm Innovation Fellowship 2025 Project! .
Jun 18, 2026	Our Paper “MED-DSLC: Multi-Expert-Domain Classification via Domain Supervision and Logit Calibration” accepted to ECCV’26! .
Feb 20, 2026	My Qualcomm Internship work “Video Reasoning Without Training” accepted to CVPR Findings 2026.
May 27, 2025	Awarded the “Qualcomm Innovation Fellowship 2025” for our proposal on Meta-Prompting! .
Sep 29, 2024	Gave an oral presentation of my PromptSliders paper at the Unlearning and Model Editing Workshop in ECCV’24 .

selected publications

2026

MED-DSLC: Multi-Expert-Domain Classification via Domain Supervision and Logit Calibration

Deepak Sridhar*, Zheng Zeng*, and Nuno Vasconcelos

In European Conference on Computer Vision (ECCV), 2026

Bib PDF

@inproceedings{sridhar2026meddslc,
  title = {MED-DSLC: Multi-Expert-Domain Classification via Domain Supervision and Logit Calibration},
  author = {Sridhar*, Deepak and Zeng*, Zheng and Vasconcelos, Nuno},
  year = {2026},
  booktitle = {European Conference on Computer Vision (ECCV)},
  primaryclass = {cs.CV},
}

Video Reasoning Without Training

Deepak Sridhar*, Kartikeya Bhardwaj*, Jeya Pradha Jeyaraj, and 3 more authors

In Computer Vision and Pattern Recognition (CVPR) Findings, 2026

Bib PDF

@inproceedings{sridhar2025vreason,
  title = {Video Reasoning Without Training},
  author = {Sridhar*, Deepak and Bhardwaj*, Kartikeya and Jeyaraj, Jeya Pradha and Vasconcelos, Nuno and Nayak, Ankita and Teague, Harris},
  year = {2026},
  booktitle = {Computer Vision and Pattern Recognition (CVPR) Findings},
  primaryclass = {cs.CV},
}

2024

Adapting Diffusion Models for Improved Prompt Compliance and Controllable Image Synthesis

Deepak Sridhar, Abhishek Peri, Rohit Rachala, and 1 more author

In Neural Information Processing Systems, 2024

Bib PDF

@inproceedings{neuripssridhar24,
  author = {Sridhar, Deepak and Peri, Abhishek and Rachala, Rohit and Vasconcelos, Nuno},
  title = {Adapting Diffusion Models for Improved Prompt Compliance   and Controllable Image Synthesis},
  booktitle = {Neural Information Processing Systems},
  year = {2024}
}

Prompt Sliders for Fine-Grained Control, Editing and Erasing of Concepts in Diffusion Models

Deepak Sridhar, and Nuno Vasconcelos

In In Proceedings of the IEEE/CVF European Conference on Computer Vision Workshops, 2024

Bib PDF

@inproceedings{eccvsridhar24,
  author = {Sridhar, Deepak and Vasconcelos, Nuno},
  title = {Prompt Sliders for Fine-Grained Control, Editing and Erasing of Concepts in Diffusion Models},
  booktitle = {In Proceedings of the IEEE/CVF European Conference on Computer Vision Workshops},
  year = {2024}
}

2023

SCHEME: Scalable Channel Mixer for Vision Transformers

Deepak Sridhar, Yunsheng Li, and Nuno Vasconcelos

2023

Bib PDF

@misc{sridhar2024schemescalablechannelmixer,
  title = {SCHEME: Scalable Channel Mixer for Vision Transformers},
  author = {Sridhar, Deepak and Li, Yunsheng and Vasconcelos, Nuno},
  year = {2023},
  eprint = {2312.00412},
  archiveprefix = {arXiv},
  primaryclass = {cs.CV},
}

2021

Class semantics-based attention for action detection

Deepak Sridhar, Niamul Quader, Srikanth Muralidharan, and 3 more authors

In In Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021

Bib PDF Supp

@inproceedings{sridhar21,
  author = {Sridhar, Deepak and Quader, Niamul and Muralidharan, Srikanth and Li, Yaoxin and Dai, Peng and Lu, Juwei},
  title = {Class semantics-based attention for action
                  detection},
  booktitle = {In Proceedings of the IEEE/CVF International
                  Conference on Computer Vision},
  year = {2021}
}