An object detection and segmentation for Oysters. It uses Detecron2 implementation.
Modify oysterd/config.py
to modify the training and evalutation data folder, the folder containing the video to be used for inference, and the network structure of the model defined in Detectron2 Model Zoo. You can assign multiple network structures here.
Either run python oysterd/train.py -i <model_number>
to train a single model <model_number>
defined in ysterd/config.py
, or run sh run_trains.sh
for training multiple models.
Use python oysterd/infer_video.py -i <model_number>
if you defined the video path in the oysterd/config.py
. You can also pass the video path by -v
argument.