Saurabh Buttan

Research & Publications

Machine learning for interaction, computer-vision research, and experimental neural rendering.

ResearchPublications

Research in machine learning & vision

Three areas of research: learning from movement, analyzing images, and reconstructing rendered scenes.

Rochester Institute of Technology
01 / RIT · MACHINE LEARNING

Learning from movement

Contributed to research at RIT’s MAGIC Center, training machine-learning models on Xbox Kinect data using cloud GPU notebooks to explore gestures, postures, and human–computer and animal–computer interaction. My contribution centered on model training and interpreting movement data as gestures and postures, connecting machine-learning experiments to interaction research.

RIT’s Interaction, Media, and Learning Lab lists me among its graduate students and alumni. View the RIT lab listing ↗

Machine learningModel trainingInteraction analysisKinect dataCloud GPU notebooks
Indian Institute of Science
02 / IISc · COMPUTER VISION

From prototype to visual-analysis tools

Prototyped algorithms in MATLAB, implemented image processing with OpenCV on Linux, and worked with Qt/C++ and JavaScript interfaces at IISc. Also contributed Hough-transform work using OpenCV, acknowledged in a UAV power-line detection paper and represented the Computational Intelligence lab at IISc Open Day in 2015 and 2016.

This work developed my skills in algorithm prototyping, image processing, and translating research into usable tools. The published work below covers classical computer vision: moving-vehicle detection under static and moving camera backgrounds, and line/word/character segmentation using eccentricity transforms.

MATLABOpenCVLinuxQt / C++Raspberry Pi
03 / NUBIX · NEURAL RENDERING

Neural reconstruction for sparse-volume rendering

Implementing and evaluating a recurrent, kernel-predicting network for noisy volume-path-traced images, with Python/PyTorch training, TensorRT inference, and live GPU validation.

Data & training
Sequence-based captures, linear-HDR references, guide-channel inputs, recurrent image history, and spatial/temporal reconstruction losses.
Deployment
Model export and TensorRT FP16 inference, connected to the renderer’s reconstruction workflow.
Research boundary
An implementation study of published sparse-volume reconstruction methods, separate from my two authored publications. Adaptive sampling remains experimental.

Publications & research contributions

2018SPRINGER
COMPUTER VISION · CONFERENCE PAPER

On-Road Moving Vehicle Detection by Spatio-Temporal Video Analysis of Static and Dynamic Backgrounds

Saurabh Buttan & Kavya Venugopal · Ambient Communications and Computer Systems, AISC 696, pp. 703–715.

Investigates moving-vehicle detection with both stationary and moving camera backgrounds. The approach combines background modelling, Hough-transform processing for dynamic scenes, three-frame differencing for static scenes, and constrained bounding boxes.

Read the publisher’s abstract & paper ↗
2016IJEEDC
IMAGE PROCESSING · JOURNAL ARTICLE

A Shape Based Text Segmentation Using Eccentricity Transform

Saurabh Buttan & Chandrakala H. T. · International Journal of Electrical, Electronics and Data Communication, Vol. 4, Issue 6, pp. 1–4.

Uses eccentricity transforms to separate document text into lines, words, and characters. The implementation extends to Raspberry Pi 2 for video processing and was tested on 30 printed English documents.

View the publication & abstract ↗
2015IEEE CATCON
RESEARCH CONTRIBUTION · ACKNOWLEDGMENT

Automatic detection of powerlines in UAV remote sensed images

2015 International Conference on Condition Assessment Techniques in Electrical Systems · pp. 17–21.

The study detects power lines in UAV imagery using K-means clustering, an unsupervised machine-learning method. The Davies–Bouldin index selects the number of clusters, and morphological operations refine the extraction.

My contribution: Hough-transform work using OpenCV, credited in the paper’s acknowledgment.

OpenCVHough transformComputer visionUAV imagery
View paper on IEEE Xplore ↗

Skills and the work behind them

Here’s where I’ve used each skill. Follow a project link for the details.

MACHINE LEARNING

Interaction research & reconstruction