top of page
White Structure

Ongoing
Projects

Vision Autonomy for Robots
3D reconstruction with fusion of SFM and Nerf
Rasterizing Solar defect masks for early defect analysis
Robot Mapping for dynamic Forest Environments
Understanding of Human Spine and implementing classification approach
Analysis of Changes in Permafrost of Artic Regions

Completed Projects

Video Analytics: Surveillance using novel face recognition 

The novel face recognition masks were developed by considering regressive points on the face. These regressive points helped in calculating certain angles that was different for every face. The key technologies involved are: a) Video to frames: Best frame from shot learning

b) Face extraction from the frame

c) Image preprocessing using image processing

d) Pose detection and face classification

e) prediction of regressive points

f) Construction of face masks and angle prediction

g) Angle matching with available database

h) Person identification and Transfer learning 

face recognition image.jpg

Accident protection system

IoT based accident protection system. Sensors used include: a) Ultrasonic sensor b) Temperature sensor c) Alcohol - gas sensor d) Face tracking camera sensor.

The project was tested on ESP32 board with communication over HTTP and thingspeak, and the project was implemented on Raspberry Pi 4.

AM_active-360.jpg

Implementation of Residual Blocks as CSI in 5G mmWave channel

The neural network predicts the channel estimation vector for a specific range of SNR value in real time.

A simple CNN is replaced with a residual block to prevent overfitting and to predict values with least minimum mean square error.

The Final model is deployed on hardware to look into real time prediction of CSI vector.

Alternative Path Planning with Turtlebot3 using Deep Learning

The architecture was built considering the disadvantages of 2D lidar scanner for extensive path planning. The ROS stack for a differential drive robot operates from the loaded map to the way-point cost map based navigation. A camera sensor is used along with the 2D Lidar and the data is processed from the camera sensor when the 2D Lidar fails. The above methodology with proper flowchart is discussed in the paper "Intelligent Navigational Tactics for Differential Drive Robots: Expanding boundaries".

Other Completed
Projects

  • Analysis of Astable mode of 555 Timer IC using LT spice

  • Automated car parking system using Arduino and C programming.

  • Soil Moisture sensing system using Arduino and python.

  • Contactless body temperature analysis and dispenser system using STM32F103C8T6 and Raspberry Pi 4.

  • IoT based pulse and body temperature measurement using esp32 and thingspeak.

  • Traffic time analysis using image processing.

  • Understanding Deep Learning applications in MIMO using MATLAB and python.

  • Accident protecting system with Raspberry Pi 4.

  • Video Analytics for hospitals: Capstone Project

  • Early detection and segmentation of Lesion in MRI images of brain.

bottom of page