"nms_thres": 0.4,
"img_size": 416,
"device": device
}
classes = load_classes(yolo_params["class_path"])
#Colors
cmap = plt.get_cmap("rainbow")
colors = np.array([cmap(i) for i in np.linspace(0, 1, 13)])
#np.random.shuffle(colors)
#
model = 'yolo'
detectron = YOLOv3Predictor(params=yolo_params)
dir = "tests"
for image in os.listdir(dir):
path = os.path.join(dir, image)
print(path)
img = cv2.imread(path)
detections = detectron.get_detections(img)
if len(detections) != 0:
detections.sort(reverse=False, key=lambda x: x[4])
for x1, y1, x2, y2, cls_conf, cls_pred in detections:
print("Item: %s, Conf: %.5f" % (classes[int(cls_pred)], cls_conf))
color = colors[int(cls_pred)]
#Classes
classes = load_classes(yolo_params["class_path"])
#Colors
cmap = plt.get_cmap("tab20")
colors = np.array([cmap(i) for i in np.linspace(0, 1, 20)])
np.random.shuffle(colors)
#Faster RCNN / RetinaNet
#model = 'faster'
#detectron = Predictor(model='retinanet',dataset= dataset, CATEGORIES = classes)
#YOLO
yolo = YOLOv3Predictor(params=yolo_params)
img = cv2.imread('galymzhan.jpeg')
#detections = detectron.get_detections(img)
detections = yolo.get_detections(img)
#print(detections)
unique_labels = np.array(list(set([det[-1] for det in detections])))
n_cls_preds = len(unique_labels)
bbox_colors = colors[:n_cls_preds]