MULTI-OBJECT DETECTION: BOAT, DOCK AND HUMAN

April 2016

Description:

  1. The goal of this project was to safegaurd the boat from the objects around it. So, a system was required which could be mounted on the boat. Once mounterd
    requires no human intervention apart from switching it on raise an alarm (LED) whenever human, boat or dock is near to the boat.

  2. The application uses the power of DL to perform this task.
  3. The dataset was provided by the client to train the network. Size of the dataset was consists of 50K training images, 10K each for validation and testing.
  4. The dataset was labelled manually.
  5. The AlexNet based on Caffe framework was used.

  6. The complete training and development was done on NVIDIA-JetsonTK1.
    This development board has NVIDIA GPU capable enough to train small networks like AlexNet.

  7. The whole application was automated using SHELL scripting on Linux platform.

Deep Learning Computer Vision
Avatar
Sumit Vaise
Graduate Research Assistant (2019-2020)
Computer Vision Engineer(2015-2018)

Graduate student at the Concordia University. Passionate about ML, DL, Computer Vision and Data Science.