Scholar
Harsh Rangwani
Google Scholar ID: OQK0WREAAAAJ
Research Scientist, Adobe Research
Citations & Impact
Citations
432
H-index
9
i10-index
9
Publications
20
Co-authors
27
list available
Publications
Resume (English only)
Academic Achievements
- Publications: 1. “DeiT-LT: Distillation Strikes Back for Vision Transformer Training on Long-Tailed Datasets” accepted to CVPR'24; 2. “Selective Mixup Fine-Tuning for Optimizing Non-Decomposable Metrics” accepted to ICLR'24 for a Spotlight (Top-5%) presentation; 3. “Inducing Smoothness Regularization in Federated Learning” accepted to WACV'24. Other academic activities: 1. Invited Talk on “Learning from Limited and Imperfect Data” at IEEE Deep-Tech Seminar; 2. Serving as a reviewer for ECCV'24, ICML'24, AAAI'24, and ICLR'24; 3. Organized the Adobe-IISc GenAI Workshop; 4. Attended ICCV'23 in Paris with travel grants from Google and Kotak AI Centre.
Research Experience
- Working at the intersection of Machine Learning and Computer Vision. Interned with Amazon in the Outbound Marketing Automation team and was the lead maintainer of the placement portal site.
Education
- PhD: Department of Computational and Data Science (CDS) at Indian Institute of Science, Advisor: Prof. Venkatesh Babu; Honors Degree: Department of Computer Science, IIT BHU Varanasi, Advisors: Dr. Rajeev Sangal and Dr. A.K. Singh.
Background
- Research Interests: Deep learning algorithms that train and generalize well on real-world datasets, especially in the presence of class imbalances and distribution shifts. Professional areas: Computer Vision, Generative Models, Long Tail Learning, Domain Adaptation of Models, Loss Landscape and Optimization.
Miscellany
- Personal interests: Reading books and playing basketball.
