Past Projects

End to End driving (2018)

  • Used Unity simulator to collect the good driving data.
  • Using input images from the three camera and corresponding steering angle trained a feed forward Convolution neural network.
  • Used the trained model to drive on different tracks in Unity simulator.
  • Project Utilized: PyTorch, Python, Unity Simulator.

Code: https://github.com/amita-kapoor/Behavioural-Cloning-SDC

Navigation (2018) 

  • Trained a Unity ML agent to navigate through the state space to collect Yellow bananas
  • Trained three layered feed forward Deep Q Network.
  • Used Replay Buffer and soft update.
  • Project Utilized: PyTorch, Python, UnityML

Facial Keypoint Detection and Real-time Filtering (2017)

  • Created and trained a CNN network to detect 15 facial keypoints.
  • The project takes in an input image/video and can detect the keypoints.
  • Project Utilized: Python, Keras, TensorFlow, OpenCV

Development of Garden-bot: 

An automatic irrigation system sensing soil moisture content using self made sensor for efficient use of resources and optimum flow of water in the fields (2016).

  • Developed an automated irrigation system developed to take care of your plants in your absence.
  • Uses self-made low-cost capacitive sensors for the detection of water level in plants.
  • Sensors are interfaced with Raspberry Pi microcontroller board. 
  • Algorithm was developed to detect the threshold moisture levels and control the water inflow for efficient use of water. 
  • Java-based app interface is also developed to allow manual control and inspection of the moisture content.