About

Rameswar Panda

Research Staff Member

Publications Research

Who they work with

Rogerio Feris

Categories

Computer Vision Machine Learning Multimodal Learning Transfer Learning

Rameswar Panda is a research staff member at MIT-IBM Watson AI Lab, where he works on solving real world problems using computer vision and machine learning. In particular, his focus is on learning with limited supervision (transfer learning, few-shot learning) and dynamic computation for several computer vision problems. He received a PhD from UC Riverside in 2018, under the supervision of Prof. Amit K. Roy-Chowdhury. During this research, he was very fortunate to have worked at NEC Labs America (Cupertino, Summer 2018), Adobe Research (San Jose, summer/fall 2017) and Siemens Corporate Research (Princeton, summer 2016).

Publications with the MIT-IBM Watson AI Lab

ConStruct-VL: Data-Free Continual Structured VL Concepts Learning

CVPR

Teaching Structured Vision & Language Concepts to Vision & Language Models

CVPR

CODA-Prompt: COntinual Decomposed Attention-based Prompting for Rehearsal-Free Continual Learning

CVPR

AnyDA: Anytime Domain Adaptation

ICLR

Learning to Grow Pretrained Models for Efficient Transformer Training

ICLR

Multitask Prompt Tuning Enables Parameter-Efficient Transfer Learning

ICLR

How Transferable are Video Representations Based on Synthetic Data?

NeurIPS

FETA: Towards Specializing Foundation Models for Expert Task Applications

NeurIPS

Selective Regression under Fairness Criteria

ICML

VALHALLA: Visual Hallucination for Machine Translation

CVPR

Task2Sim: Towards Effective Pre-training and Transfer from Synthetic Data

CVPR

Can an Image Classifier Suffice For Action Recognition?

ICLR

RegionViT: Regional-to-Local Attention for Vision Transformers

ICLR

IA-RED^2 : Interpretability-Aware Redundancy Reduction for Vision Transformers

NeurIPS

Dynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data

NeurIPS

Contrast and Mix: Temporal Contrastive Video Domain Adaptation with Background Mixing

NeurIPS

AdaMML: Adaptive Multi-Modal Learning for Efficient Video Recognition

ICCV Multimodal Learning

Dynamic Video Quantization for Efficient Inference

ICCV ICCV

A Broad Study on the Transferability of Visual Representations With Contrastive Learning

ICCV

CrossViT: Cross-Attention Multi-Scale Vision Transformer for Image Classification

ICCV

Detector-Free Weakly Supervised Grounding by Separation

ICCV

Multimodal Clustering Networks for Self-Supervised Learning From Unlabeled Videos

ICCV

Deep Analysis of CNN-based Spatio-temporal Representations for Action Recognition

CVPR Computer Vision

NASTransfer: Analyzing Architecture Transferability in Large Scale Neural Architecture Search

AAAI Computer Vision

Non-Adversarial Video Synthesis with Learned Priors

CVPR Computer Vision

Camera On-boarding for Person Re-identification using Hypothesis Transfer Learning

CVPR Computer Vision

Semi-Supervised Action Recognition with Temporal Contrastive Learning

CVPR Computer Vision

Fair Selective Classification Via Sufficiency

ICML

AdaFuse: Adaptive Temporal Fusion Network for Efficient Action Recognition

ICLR Computer Vision

VA-RED^2: Video Adaptive Redundancy Reduction

ICLR Computer Vision

AdaShare: Learning What To Share For Efficient Deep Multi-Task Learning

Computer Vision Multimodal Learning

AR-Net: Adaptive Frame Resolution for Efficient Action Recognition

ECCV Computer Vision

Relationship Matters: Relation Guided Knowledge Transfer for Incremental Learning of Object Detectors

CVPR Computer Vision