from darkflow.net.build import TFNet
import cv2, glob
options = {
"model": "cfg/etablev2.cfg",
"load": "bin/cust_ev2_30000.weights",
"threshold": 0.6,
"labels": "cfg/etable.names"
}
# options = {"model": "cfg/yolov2.cfg", "load": "bin/yolov2.weights", "threshold": 0.4, "labels": "cfg/coco.names"}
tfnet = TFNet(options)
for img_path in glob.glob("sample_img/*.JPG"):
imgcv = cv2.imread(img_path)
result = tfnet.return_predict(imgcv)
for box in result:
imgcv = tfnet.drawdict(imgcv, box)
cv2.imwrite("res/" + img_path.split('/')[-1], imgcv)
