RGB Infrared Cross Modal Person Re-identification

  1. Extending the idea of Rgb-Rgb Person re identification for Rgb-Ir images, a very importantuse case for night time surveillance systems.
  2. Trained a GAN based Image to Image translation model to disentangle an image into its content andstyle components in latent space.

Disentangling hidden factors of variations in Deep Networks

  1. Trained an autoencoder to disentangle the label independent features and label discriminative features in the latent space, thereby increasing the robustness of the downstream task by filtering out the nondiscriminative features.
  2. Experiments were carried out on Japanese female face expression dataset to disentangle the discriminative expression features and the non discriminative subject face features in the latent space.
  3. Efficiency of the method was measured qualitatively by swapping the face expressions across the subject faces.

Object Search in Videos

  1. Implemented a technique to effectively search a query object in the video frames.
  2. Harris Laplace corner points described using SIFT descriptors were used as features which were then vector quantised into K clusters using K means clustering.
  3. Bag of words model was used to represent each frame as a Term Frequency- Inverse document frequency (TF-IDF) vector, which was then used to calculate the cosine similarity with the query vector to rank the frames.

Re-imagination of Digital Product Store

  1. Digitization of Telecom retail store to minimize the human interference in its operations.
  2. Development of web application to be run on a Digital kiosk, to ease the sim card purchase activity for customers.
  3. Designed Recommendation algorithm to analyze the user’s past usage trends. Accordingly the perfect SMS, calling and Internet schemes are recommended to the user.

Software defined Cognitive Radio

Implemented a Software defined Cognitive Radio in MATLAB. Cognitive radio aims at effective spectrum utilization using dynamic spectrum allocation involving 3 steps:

  1. Spectrum sensing- To find vacant spectrum
  2. Spectrum allocation
  3. Measurement and modelling of spectrum usage

VLSI Chip design projects

  1. Completed schematic entry, layout and corner simulations (with pad frame) of SRAM, Carry look ahead adder using AMI 0.5 micron technology.
  2. Real time image processing using FPGA Spartan 3E.

Embedded System Projects

  1. Designed a custom BLE profile to control various digital appliances using a given Android application. Project was designed using PSoC Creator and implemented on Cypress PSoC4, Pioneer Board.
  2. Implemented a sensor node that could sense data from any connected sensor and transmit it over an Ethernet interface to a Sparkfun Phant Server / Project was designed using ARM mbed platform, and implemented on Freedom Board (FRDM KL25Z).
  3. Built a prototype of Home automation project to control the home appliances using Bluetooth HC 05 as communication interface. Project was implemented using Atmel’s AVR Atmega 8 microcontroller.

© 2018. All rights reserved.

Powered by Hydejack v8.4.0