This is our term project for course COMP6733 IoT Design Studio (22T3) @ UNSW.
SmartEdge Tracker is a realtime GUI-based(touch monitor friendly) object tracking application developed for NVIDIA Jetson Nano.
This application should also work on other NVIDIA Jetson devices, but we haven't tested yet as we don't have other Jetson devices.
For this particular project, we used the dataset from Kaggle and trained a yolov5n(YOLOv5 v6.0) model detecting Crown-of-Throns Starfish(COTS).
We also added a SORT object tracker to track and count the number of COTS.
If you want to track other objects, you can also train your own model (yolov5n), convert the weights to .wts and replace the model file in the weights directory.
The application will then automatically create a .engine file (TensorRT optimized model) for inference.
Below is a screenshot of the application. You can also find our demo video HERE on YouTube.
Follow the instructions in this repo jetson-setup.
Run with camera: python3 main.py
Run with video file: python3 main.py <video file path>.
The demo video can be downloaded from GoogleDrive
If the camera failed to load frames, close the app, execute
sudo systemctl restart nvargus-daemon
, and re-launch the demo script.
CSIRO cooperated with Google developed a real life application in a larger scale: Using Machine Learning to Help Protect the Great Barrier Reef in Partnership with Australia’s CSIRO. Here's a rough comparison between our system and Google's solution(one slide extracted from our project presentation):
I am impressed very much by the the quality of the projects. I would like to particularly highlight the following project and team. "AI On Edge: Help Protect the Great Barrier Reef" by Team SmartEdge (Members' names are omitted due to privacy concerns). The guest lecturer from CSRIO that works on the project is also very impressed by the results achieved.