avatar

Aniket Dashpute

Ph.D. Student
Rice University
aniket.d (at) rice.edu


About Me

Hello! I am a Ph.D. student in the Computational Imaging Lab at Rice University. I am advised by Prof. Ashok Veeraghavan. I also work closely with Prof. Florian Willomitzer from University of Arizona and Prof. Oliver Cossairt from Northwestern/Meta. Previously at Northwestern University, I worked with Prof. Oliver Cossairt, Prof. Aggelos Katsaggelos, and Prof. Emma Alexander.

I graduated in Electrical Engineering from IIT Delhi. During my time there, I worked with Prof. Kedar Khare, Prof. Mukul Sarkar and Prof. Prem Kalra. After my graduation, I worked in Samsung Health, Bangalore on the rendering engine for Ultrasound modalities.

Research Interests

My research lies at the intersection of computer graphics, computer vision and computational imaging. My current research interests include neural scene representation and rendering, thermal imaging and event sensors imaging

Projects

3D Rendering Engine, Ultrasound Modalities, SAMSUNG Healthcare

Rendered images from Ultrasound data
  • Spatio-Temporal Image Correlation (STIC) Cine play - improved rendering speed by almost 10x
  • Realistic and Translucency Volume and surface rendering algorithms
  • Enhancement and Optimization of various Image Processing Filters by using Compute Shader to implement them on GPU

  • Propagation of Vector Beam through Random Media, IIT Delhi

    Rendered images from Ultrasound data
  • Devised a novel method of using a robust engineered beam to give better real time imaging through random scattering media. Employed Laguerre-Gaussian beam modes in cross polarization to generate the beam
  • Conducted extensive simulations modeling the thick diffuser and observing the quality of the combined beam after propagation
  • Performed experiments in lab and proved through SNR measurement that the combined beam had better beam retention compared to the widely used Gaussian beam
  • Combined beam was found to be robust to variations caused by scattering and time varying phase fluctuations due to the complementary nature of its component LG beams

  • Scene Adaptive Cosmetic Makeup Transfer, IIT Delhi

    Rendered images from Ultrasound data
  • Developed a system that can make-up and other accessories on face image using a reference image as guide
  • 3D morphable models used to transfer makeup from an example onto subject image to get back 2D image
  • Work is also under progress where we will add accessories in 3D first so that they blend perfectly in the scene and look natural when the resulting 2D image is obtained
  • Also implemented an existing old version of the problem solution - Digital Face Makeup by Example to see results

  • Depth Resolution Enhancement in Time-of-Flight Cameras, IIT Delhi

    Rendered images from Ultrasound data
  • Presented a novel method for reducing depth measurement error using polarization state of reflected light
  • Transmission angle of the linear polarizer was varied for a fixed object distance from camera. Sinusoidal variations were observed for depth error with clear maxima and minima
  • It was observed that the depth error was minimum when polarization angle was equal to ToF phase shift
  • Experimental results showed error reduction of upto 87% for 3000mm distance

  • Publications

    1. CVPR
      Aniket Dashpute, Vishwanath Saragadam, Emma Alexander, Florian Willomitzer, Aggelos Katsaggelos, Ashok Veeraraghavan, Oliver Cossairt
      IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023.


    Powered by Jekyll and Minimal Light theme.