[YOLO] Darkflow 사용해보기

 

// 가상환경 실행 시켜 주고 darkflow 클론 

(venv) odoaui-MacBookPro ~ % cd project

(venv) odoaui-MacBookPro project % pip install cython

(venv) odoaui-MacBookPro project % git clone https://github.com/thtrieu/darkflow

(venv) odoaui-MacBookPro project % cd darkflow

 

//Build 해주기 

(venv) odoaui-MacBookPro darkflow % python3 setup.py build_ext --inplace

 

 

//Weights model 다운로드 후 

pjreddie.com/media/files/yolov2.weights

pjreddie.com/media/files/yolov3.weights

pjreddie.com/media/files/yolov2-tiny.weights

pjreddie.com/media/files/yolov3-tiny.weights

 

 

//darkflow  에 bin 에 folder 를 만들고

/ bin 안에  Weight 파일 복사 

 

yolov3-tiny.weights 를 h5 파일로 변환 

(venv) odoaui-MacBookPro keras-yolo3 % python convert.py yolov3-tiny.cfg yolov3-tiny.weights model_data/yolo_tiny.h5

 

get session error 나서 케라스 버전 올려주고 , 텐서플로우 버전은 내려줌 

(venv) odoaui-MacBookPro keras-yolo3 % pip install keras==2.3.0

(venv) odoaui-MacBookPro keras-yolo3 % pip install tensorflow==1.14.0

 

 

jupyter notebook 실행

from IPython.display import display
from PIL import Image
from yolo import YOLO
 
def objectDetection(file, model_path, class_path):
    yolo = YOLO(model_path=model_path, classes_path=class_path, 
                anchors_path="model_data/tiny_yolo_anchors.txt")
    image = Image.open(file)
    result_image = yolo.detect_image(image)
    display(result_image)
 
objectDetection('model_data/dogcat.jpg', 'model_data/yolo_tiny.h5', 'model_data/coco_classes.txt')