Milind G. Padalkar

Since April 2025, I have been working as a Technical Lead at Infocusp Innovations.

From January 2019 to March 2025, I was a postdoctoral researcher at the PAVIS research line of the Istituto Italiano di Tecnologia (IIT), Genova, Italy, under the supervision of Dr. Alessio Del Bue (and formerly with Prof. Vittorio Murino). My research focused on computer vision and machine learning for defect detection, as well as multi-exposure and multi-illumination image fusion. I contributed primarily to industrial projects aimed at defect enhancement and detection in combustion chamber tiles, textile yarn, and infrastructure components.

Prior to joining IIT, I worked as a senior research engineer at Vehant Technologies (Mar. 2017 - Dec. 2018).

I completed my Ph.D. in 2017 under the supervision of Prof. Manjunath V. Joshi at DA-IICT, Gandhinagar, India. Before that I did my M.Tech. in 2010 under the supervision of Prof. M. A. Zaveri at SVNIT, Surat, India and B.E. in 2008 from FAMT, Ratnagiri (The University of Mumbai).

I have co-authored the book titled "Digital Heritage Reconstruction Using Super-resolution and Inpainting" published under Synthesis Lectures on Visual Computing by Morgan & Claypool Publishers in December 2016.

LinkedIn  /  Google Scholar  /  Twitter  /  Github  /  CV

The paper presents a teacher-student network that hallucinates 3D features directly from RGB images to improve the keypoints localization.

Various defects in bridges are detected using patch based classifiers.

Our paper presents an end-to-end approach that leverages single-view RGB images for estimating an ordered list of 3D key-point.

Our paper explains how a hybrid image fusion technique can help to enhance the human annotation performance for industrial defect detection.

Crack details are combined from several mutually registered multi-illumination images of ceramic tiles, into a single representative image.

A novel setup for automatic visual inspection of cracks in ceramic tile as well as studies the effect of various classifiers and height-varying illumination conditions for this task.

Our book presents image super-resolution methods and techniques for automatically detecting and inpainting damaged regions in heritage monuments, in order to provide an enhanced visual experience.

In this chapter, we discuss techniques for automatically detecting the damaged facial regions and cracks in photographs of monuments. Unlike the usual practice of manually selecting the mask for inpainting, the regions to be inpainted are automatically selected and inpainting is done using the existing algorithm.

We construct dictionaries of image-representative low and high resolution patch pairs from the known regions in the test image and its coarser resolution. Inpainting of the missing pixels is performed using exemplars found by comparing patch details at a finer resolution by making use of the constructed dictionaries.

A technique for automatically detecting the cracked regions in photographs of monuments, based on comparison of patches using a measure derived from the edit distance. This is extended to perform inpainting of video frames using SIFT and homography, which is quantified with our temporal consistency measure.

Automates the process of identifying the damage to visually dominant regions viz. eyes, nose and lips in facial image of statues, for the purpose of inpainting.

We present a Singular Value Decomposition (SVD) based technique for automatic detection of the damaged regions in the photographed object/scene, for digitally restoring the entirety using inpainting.

Unlike simply copying pixels from an exemplar into the target or damaged pixels (i.e. pixels to be inpainted), we estimate and use autoregressive parameters along with the best matching exemplar to modify the damaged pixel.

A dual stage divide-and-merge approach to detect video-shots joined by dissolve type transitions.

This is an old code that I created sometime in 2010. While working on video shot detection, I needed a tool to perform frame-by-frame visual analysis. However, I wasn't aware of any that existed at that time and created one with Matlab. To sharpen my skills with C and OpenCV, I re-created the same.

Website built using source code provided by Jon Barron.

Last updated on: