Publications | Karttikeya Mangalam

Reversible Vision Transformers

MeMViT: Memory-Augmented Multiscale Vision Transformer for Efficient Long-Term Video Recognition

Improved Multiscale Vision Transformers for Classification and Detection

From Goals, Waypoints & Paths To Long Term Human Trajectory Forecasting

Multiscale Vision Transformers

LOKI: Long Term and Key Intentions for Trajectory Prediction

Overcoming Mode Collapse with Adaptive Multi Adversarial Training

Object-Region Video Transformers

It Is Not the Journey but the Destination: Endpoint Conditioned Trajectory Prediction

Long-term Human Motion Prediction with Scene Context

Disentangling Human Dynamics for Pedestrian Locomotion Forecasting with Noisy Supervision

Learning Spontaneity to Improve Emotion Recognition in Speech

Future Person Localization in First-Person Videos

Do deep neural networks learn shallow learnable examples first?

Bitwise Operations of Cellular Automaton on Gray-scale Images