This repo share scripts and code used in the Baltic Seabird Project.
The seabird detector is basically a wrapper around the yolov5 object detector.
Note on the data...
Clone the repo
git clone https://github.com/BalticSeabird/SeabirdDetection.git
cd SeabirdDetection
Build the docker image
sh build/build.sh
First adjust the 'DATA_DIR' in 'train.sh' to reflect your setup and set the image size and model in 'train.py'. Then start the training
sh train.sh
In order to improve the performance of the inference on gpu you can export the model to TensorRT
sh export.sh