<<<<<<< HEAD
As we all know, queen bee is relatively hard to be detected among a giant group of worker bees, because of the similar appearance and their tiny size for human beings. So if we wanna detect the queen bee, there are a lot of tedious things to do, including construct the datasets with the queen bee contained, then we should utilize an efficient detection model which could be fine-tuned to the novel datasets. About how to distinct the queen bee from worker bees, the dominant help is the relative size between the queen and it's surrounding worker bees, concretely elaborated in this blog:
Using the SOTA object detection model: yolov8. It's pipeline is shown below:
Within YOLOv8, we have different model sizes such as yolov8- n - nano, s - small, m - medium,
l - large, and x - extra large.
In the project, I select three types of yolov8 models as the base model which are pretrained on large datasets , .i.e, MicroSoft COCO or Imageset, users can select one of which on the basis of the performance of devices.
conda create -n yolo_env python==3.8
conda activate yolo_env
conda install pytorch torchvision torchaudio -c pytorch
pip install -U ultralytics
Run the finetune_training.py, the yolov8 model will be trained under the queen_worker dataset.
python finetune_training.py
After getting trained model, we can use this model to detect an image or a video.
# test img
result = model(img_path, save=True, project="results/")
# test video
result = model(video_path, save=True, project="results/")
At last, if we run DetQueen.py, we will see an user interface in which we can
input images and videos, and waiting for a while, the corresponding results will be
demonstrated.
From a group of worker bees, find the queen bee
e87345269bc67bce0f56b7883b28552306ce2538