Research in engineering has a broad perspective. It includes establishment of theoretical concepts, new inventions, innovative alternates to the existing solutions, development of new techniques and systems, and much more. It is a journey in which a researcher achieves milestones, deals with setbacks and grows with experience. My journey began in 2009 with a summer project on watermark detection techniques. Subsequently, I became part of the healthcare research and development team in Siemens Information Systems Ltd. I was fortunate to come into contact of different faculty members and mentors who provided me the opportunities to work on challenging problems. It setup the platform for an in-depth and rigorous research and motivated me to join IIT Delhi for higher studies. IIT Delhi opened up several doors with tremendous possibilities. I explored different domains while consistently remaining active in my doctoral research, which has been published in elite journals and conferences. Currently I am an Assistant Professor at the Department of Computer Science and Engineering, IIT Jodhpur.
As “an intrigued mind never sleeps”, while remaining studious, I am actively looking for motivated PhD students who are interested to join me in the areas of machine learning, resource constrained AI, dependable AI, medical image analysis, and hardware design.
PhD in Computer Technology (Electrical Engineering), 2019
Indian Institute of Technology Delhi
M. Tech in Computer Technology (Electrical Engineering), 2013
Indian Institute of Technology Delhi
B. Tech in Electronics and Communication Engineering, 2010
Indian Institute of Information Technology Jabalpur
“The only time you mustn’t fail is the last time you try." - Charles Kettering
Highlights of my research are as follows.
Sub-problem specific deep supervision: FCNN/CNNs are used to infer high-level context using low-level image features. In this paper, a sub-problem specific deep supervision of the FCNN is performed. The attention of fine resolution layers is steered to learn object boundary definitions using auxiliary losses, whereas coarse resolution layers are trained to discriminate object regions from the background. Furthermore, a customized scheme for downweighting the auxiliary losses and a trainable fusion layer are introduced. This produces an accurate segmentation and helps in dealing with the broken boundaries, usually found in the ultrasound images.
Structure oriented GAN: In this work we proposed a deep adversarial despeckling approach to enhance the quality of ultrasound images. Most of the existing approaches target a complete removal of speckle, which produces oversmooth outputs and results in loss of structural details. In contrast, the proposed approach reduces the speckle extent without altering the structural and qualitative attributes of the ultrasound images. A despeckling residual neural network (DRNN) is trained with an adversarial loss imposed by a discriminator. The discriminator tries to differentiate between the despeckled images generated by the DRNN and the set of high-quality images. Further to prevent the developed network from oversmoothing, a structural loss term is used along with the adversarial loss.
Blood oxygen saturation measurement using optical sectioning: A linear relationship between the blood oxygen saturation and ratio of the partially polarized and polarized light components reflected from fingertip is observed. It is used to develop a portable low cost system for non-invasive and non-contact blood oxygen saturation measurement.
Bio-image computing: Bio-image computing is an upcoming area with incredible possibilities. It has paramount importance for a country like India, which greatly depends on the agriculture industry. In our work, we showed that electron microscopy (EM) image based quantification of stomatal opening and stomata counting can provide an autonomous tool to measure the effect of environmental stress on the plants.
EPPRSRAD: Anisotropic diffusion filters are one of the best choices for speckle reduction in the ultrasound images. These filters control diffusion flux flow using local image statistics and provide the desired speckle suppression. However, inefficient use of edge characteristics results in either oversmooth image or an image containing misinterpreted spurious edges. As a result, diagnostic quality of the images becomes a concern. To alleviate such problems, an edge probability and pixel relativity based anisotropic diffusion speckle reducing filter is developed. The probability density function helps in removing spurious edges and the pixel relativity reduces oversmoothing effects. Furthermore, the filtering is performed in superpixel domain to reduce the execution time, wherein a minimum of 15% of the total number of image pixels can be used.
Other relevant publications