|
I am currently a Principal Engineer at Akasha Imaging, where I am also a founding team member. My research interests lie at the intersection of computational imaging, multi-view geometry, and machine learning, especially in the context of visual diversity or robotics. I have published in NeuriPS, ICCV, and CVPR and am a co-inventor on more than a dozen patents in computational imaging.
|
|
|
Update August 28th, 2021: Excited to be giving a SCIEN talk on October 13th at Stanford! Update July 22nd 2021: Our new paper on rotation invariance was accepted to ICCV 2021! Stay tuned... |
|
|
|
Agastya Kalra, Vahe Taamazyan, Supreeth Krishna Rao, Kartik Venkataraman, Ramesh Raskar Achuta Kadambi CVPR, 2020 (Oral Presentation) paper / supplement / teaser / video / demo We leverage polarization cues in neural pipelines to improve invarances w.r.t. transparent objects in segmentation and robotic bin picking |
|
Abdullah Rashwan, Agastya Kalra, Pascal Poupart ICCV workshop, 2019 (Best Paper Shortlist) Short Paper / Long Paper / Code We re-architect the Feature Pyramid Networks to improve invariance to aspect ratios and show SOTA performance in COCO test and val. |
|
Agastya Kalra, Ben Peterson arXiv, 2019 Paper / Demo / Article / The Batch We leverage deep voter modeling to rate dating photos and allow millions of users to select the best dating photo. |
|
Agastya Kalra, Abdullah Rashwan, Wilson Hsu, Pascal Poupart, Prashant Doshi, Georgios Trimponias NeurIPS, 2018 Paper / Code / Video We propose the first successful online structure learning method for gaussian and recurrent Sum-Product Networks. |
|
|
|
Guest Lecturer and Course Advisor, CS480 (Machine Learning) Fall 2017 with Yaoliang Yu.
Guest Lecturer and Course Advisor, CS484 (Computer Vision) Fall 2018 with Yuri Boykov. Guest Lecturer in Computer Vision and Machine Learning courses with Pascal Poupart, Yuri Boykov, Yaoliang Yu, and Jeff Orchard. |
|
Template credits: Jon Barron |