Tournament name: 'Drone Surveillance Contest'.
Tournament task: to count and track the number of cars by drone passing over a junkyard.

Drone Surveillance Counter

Project steps:
1. Create an image dataset
* Split the video into frames
* Add new images data to prevent overfitting and achieve versatile performance
* Perform augmentation to achieve the optimum result for a dynamic environment
* Data labeling
2. Training custom CNN model (YOLO Algorithm)
3. Code
* Initialize Model
* Cars Detection
* Count and track cars that are only in the region of interest.
* Draw result

Video processing by CUDA technology support, NVIDIA® GeForce® graphic card.

Leave a Reply

Your email address will not be published.