I am a research scientist at Nvidia Research in Westford, US. Prior to joining Nvidia, I completed my PhD at Perceiving Systems department, Max-Planck Institute (MPI) for Intelligent Systems in Tübingen, Germany. My PhD advisor was Prof. Peter V. Gehler. I did my bachelors and masters in Computer Science at IIIT-Hyderabad, India.

My work lies at the intersection of Computer Vision and Machine Learning. Specifically, I am working on leveraging machine learning techniques for better inference in computer vision models. The main research question is how to make use of learning techniques such as deep neural networks and random forests for inference in structured prediction frameworks.

I am looking for motivated students for collaborations and internships. If interested, please drop me an email with your CV.

News

September, 2018 : Code for ‘Superpixel Sampling Networks’ is released.

August, 2018 : Code for ‘Learning Superpixels with Segmentation-Aware Loss’ is released.

July, 2018 : Papers on ‘Superpixel Sampling Networks’ and ‘Switchable Temporal Propagation Networks’ accepted to ECCV’18.

June, 2018 : Our work on SPLATNet received best paper honorable mention at CVPR’18.

June, 2018 : Code for ‘SPLATNet’ is released.

April, 2018 : Received Outstanding Reviewer award at CVPR’18.

April, 2018 : A paper on training deep networks with synthetic data is accepted to ‘Workshop on autonomous driving’ at CVPR’18.

February, 2018 : Papers on point-cloud CNNs (oral), multi-frame video interpolation (spotlight) and superpixels with deep features are accepted to CVPR’18.

see all news

varunjampani@gmail.com
Nvidia Research
2 Technology Park Drive
01886, Westford, USA.

Publications

Superpixel Sampling Networks

An end-to-end trainable deep superpixel algorithm that allows learning with flexible loss functions resulting in the learning of task-specific superpixels.

V. Jampani, D. Sun, M-Y. Liu, M-H. Yang and J. Kautz

European Conference on Computer Vision, ECCV’18

pdf / poster / video / project page / code (github)

Switchable Temporal Propagation Network

A generic fast deep learning framework for propagating a variety of visual properties (such as color, HDR, segmentation) across video frames.

S. Liu, G. Zhong, S. D. Mello, J. Gu, V. Jampani, M-H. Yang and J. Kautz

European Conference on Computer Vision, ECCV’18

pdf / code (coming soon)

SPLATNet: Sparse Lattice Networks for Point Cloud Processing

A fast and end-to-end trainable neural network that directly works on point clouds and can also do joint 2D-3D processing.

H. Su, V. Jampani, D. Sun, S. Maji, E. Kalogerakis, M-H. Yang and J. Kautz

Computer Vision and Pattern Recognition, CVPR’18 (oral, best paper honorable mention)

pdf / poster / video / CVPR talk / code (github)

Super SloMo: High Quality Estimation of Multiple Intermediate Frames for Video Interpolation

Variable-length multi-frame video interpolation via self-supervised optical flow estimation and occlusion reasoning.

H. Jiang, D. Sun, V. Jampani, M-H. Yang, E. Learned-Miller and J. Kautz

Computer Vision and Pattern Recognition, CVPR’18 (spotlight)

pdf / supplementary / news / CVPR talk / video results

Learning Superpixels with Segmentation-Aware Affinity Loss

A learning technique that leverages deep networks to predict pixel affinities useful for graph-based superpixel segmentation.

W-C. Tu, M-Y. Liu, V. Jampani, D. Sun, S-Y. Chien, M-H. Yang and J. Kautz

Computer Vision and Pattern Recognition, CVPR’18

pdf / project page / code (github)

Training Deep Networks with Synthetic Data: Bridging the Reality Gap by Domain Randomization

A study exploring the use of domain randomization for real-world computer vision tasks such as object detection.

J. Tremblay*, A. Prakash*, D. Acuna*, M. Brophy*, V. Jampani, C. Anil, T. To, E. Cameracci, S. Boochoon and S. Birchfield (*equal contribution)

Workshop on Autonomous Driving, CVPR-Workshops’18

pdf

Semantic Video CNNs through Representation Warping

A fast and end-to-end trainable approach for converting image CNNs to video CNNs for semantic segmentation.

R. Gadde, V. Jampani and P. V. Gehler

International Conference on Computer Vision, ICCV’17 (oral)

pdf / video / code (github) / ICCV talk / poster

Video Propagation Networks

A fast, generic and end-to-end trainable approach for propagating information across video frames.

V. Jampani, R. Gadde and P. V. Gehler

Computer Vision and Pattern Recognition, CVPR’17

pdf / supplementary / project page / code (github) / poster

Efficient 2D and 3D Facade Segmentation using Auto-Context

A fast auto-context based facade segmentation approach for segmenting both 2D images and 3D point clouds.

R. Gadde*, V. Jampani*, R. Marlet and P. V. Gehler (*equal contribution)

Pattern Analysis and Machine Intelligence, PAMI’17

arXiv pre-print / code (bitbucket)

Superpixel Convolutional Networks using Bilateral Inceptions

An inception style fast bilateral filtering module that can be used in existing image segmentation CNNs.

R. Gadde*, V. Jampani*, M. Kiefel, D. Kappler and P. V. Gehler (*equal contribution)

European Conference on Computer Vision, ECCV’16

pdf / supplementary / project page / code (github) / poster

Learning Sparse High-Dimensional Filters: Image Filtering, Dense CRFs and Bilateral Neural Networks

Parameterization and learning of sparse high-dimensional filters with wide range of applications in image filtering, CRFs and neural networks.

V. Jampani, M. Kiefel and P. V. Gehler

Computer Vision and Pattern Recognition, CVPR’16

pdf / supplementary / project page / code (github) / poster

Optical Flow with Semantic Segmentation and Localized Layers

An approach for incorporating semantics of the scene for better optical flow estimation.

L. Sevilla, D. Sun, V. Jampani and M. J. Black

Computer Vision and Pattern Recognition, CVPR’16

pdf / video / project page / code (zip)

Permutohedral Lattice CNNs

CNNs with high-dimensional permutohedral kernels instead of standard spatial kernels.

M. Kiefel, V. Jampani and P. V. Gehler

International Conference on Learning Representations, ICLR Workshops’15

pdf

Consensus Message Passing for Layered Graphical Models

Improved message passing inference in layered vision models that can be used in probabilistic programs like Infer.NET.

V. Jampani*, S. M. A. Eslami*, D. Tarlow, P. Kohli and J. Winn (*equal contribution)

Artificial Intelligence and Statistics, AISTATS’15

pdf / supplementary / arXiv

Efficient Facade Segmentation using Auto-Contex

A state-of-the-art approach for facade segmentation that is faster than most existing approaches.

V. Jampani*, R. Gadde* and P. V. Gehler (*equal contribution)

Winter Conference on Applications of Computer Vision, WACV’15

pdf / supplementary / code (bitbucket) / project page

The Informed Sampler: A Discriminative Approach to Bayesian Inference in Generative Computer Vision Models

A generic MCMC sampling technique for inverting graphic engines.

V. Jampani, S. Nowozin, M. Loper and P. V. Gehler

Computer Vision and Image Understanding Journal, CVIU’15

pdf / arXiv

Assessment of Computational Visual Attention Models on Medical Images

Analysis and extensions of visual attention models for medical image perception.

V. Jampani, Ujjwal, J. Sivaswamy and V. Vaidya

Indian Conference on Computer Vision, Graphics and Image Processing, ICVGIP’12

pdf / online

Role of Expertise and Contralateral Symmetry in the Diagnosis of Pneumoconiosis: An Experimental Study

An experimental eye-tracking study on X-ray image perception, with observers ranging from novices to experienced radiologists.

V. Jampani, V. Vaidya, J. Sivaswamy and K. L. Tourani

SPIE, Medical Imaging’11

online

ImageFlow: Streaming Image Search

A novel image search user interface that explores a different alternative to the traditional grid interfaces.

V. Jampani, G. Ramos and S. Drucker

Microsoft Research Technial Report, Redmond, MSR-TR-2010-148

pdf / news article


see all publications

Theses

Learning Inference Models for Computer Vision

Learning based techniques for better inference in several computer vision models ranging from inverse graphics to freely parameterized neural networks.

V. Jampani

PhD Thesis, MPI for Intelligent Systems and University of Tübingen, December, 2016

pdf / slides / library

A Study of X-Ray Image Perception for Pneumoconiosis Detection

Eye tracking experimental studies and models for X-Ray Image Perception and Diagnosis.

V. Jampani

Master Thesis, IIIT-Hyderabad, January, 2013

pdf