WACV 2026
Synthesizing Compositional Videos from Text Description
Composes multi-object videos from a text prompt with no training, on a frozen video diffusion model.
Research
Generative vision without retraining
Training-free and test-time optimization methods that keep frozen diffusion models consistent, for camera-controlled multi-view synthesis and compositional video generation.
3D scene modeling
Neural graphics primitives that learn many scenes one after another inside a single model, without catastrophic forgetting.
Decoding the brain
Generative architectures that reconstruct still and moving visual stimuli from EEG, with a focus on latent space alignment and semantic analysis.
I am also interested in representation learning, graph neural networks, multimodal learning, neural rendering and brain-computer interfaces.
Selected publications
A few papers that show where my work is headed.
Preprint 2025
Beyond Reconstruction: What EEG-to-Video Decoding Actually Recovers
Asks what EEG-to-video decoding actually recovers of visual experience.
BMVC 2025
Incremental Multi-Scene Modeling via Continual Neural Graphics Primitives
Learns many 3D scenes in a single NeRF without forgetting the earlier ones.
BMVC 2026
Why Multi-Source Audio Fails to Compose: A Geometric Diagnosis and Its Remedy
Explains why mixing several sound sources breaks image generation, and fixes it.
WACV 2024
Learning Robust Deep Visual Representations from EEG Brain Recordings
Learns robust visual representations from EEG. Featured in WACV Daily and Best of WACV 2024.
ICASSP 2023
EEG2IMAGE: Image Reconstruction from EEG Brain Signals
Reconstructs images from EEG signals with contrastive learning and a conditional GAN.
Education
PhD in Computer ScienceOct 2021 to May 2026
PMRF scholar, advised by Dr. Shanmuganathan Raman. CGPA 9.71/10.
Thesis: Scalable Generative Vision and Neural Decoding: Connecting Machine and Human Perception
Thesis: Scalable Generative Vision and Neural Decoding: Connecting Machine and Human Perception
M.Tech. in Computer Science and EngineeringJul 2019 to May 2021
Advised by Dr. Shanmuganathan Raman. CGPA 8.89/10.
Thesis: Generating 3D Shapes and Images from the Embedding of Images and Text
Thesis: Generating 3D Shapes and Images from the Embedding of Images and Text
Recognition
2024Learning Robust Deep Visual Representations from EEG Brain Recordings featured in WACV Daily and Best of WACV 2024.
2023Best Poster Award at Pacific Graphics for Hand Shadow Art: A Differentiable Rendering Perspective.
2021 to 2026Prime Minister's Research Fellowship (PMRF).
2019GATE CS: 98.69 percentile, rank 1,309 out of 99,932 candidates.
Experience
NowPostdoctoral fellowCVIG Lab, IIT Gandhinagar, India
May 2025 to Oct 2025PhD research internChat3D, France (online)
May 2024 to Nov 2024PhD research internAdobe, Bangalore, India
May 2021 to Sep 2021Junior research fellowIIT Gandhinagar, India
May 2019 to Jun 2019Freelancer, GATE CSEGradeup (Gradestack Learning Pvt Ltd)