GLUONCV.py

import argparse
import time

import gluoncv as gcv
gcv.utils.check_version('0.4.0')
from gluoncv.utils import try_import_cv2
cv2 = try_import_cv2()
import mxnet as mx

parser = argparse.ArgumentParser(description="Webcam object detection script",
                                 formatter_class=argparse.ArgumentDefaultsHelpFormatter)

parser.add_argument('--num-frames', type=int, default=200,
                    help='number of frames to run the demo for. -1 means infinite')


args = parser.parse_args()

# Load the model
net = gcv.model_zoo.get_model('ssd_512_mobilenet1.0_coco', pretrained=True)

# Load the webcam handler
cap = cv2.VideoCapture(0)
time.sleep(1)  ### letting the camera autofocus

NUM_FRAMES = args.num_frames
i = 0
while i < NUM_FRAMES or NUM_FRAMES == -1:
    i += 1

    # Load frame from the camera
    ret, frame = cap.read()

    # Image pre-processing
    frame = mx.nd.array(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)).astype('uint8')
    rgb_nd, scaled_frame = gcv.data.transforms.presets.ssd.transform_test(frame, short=512, max_size=700)

    # Run frame through network
    class_IDs, scores, bounding_boxes = net(rgb_nd)

    # Display the result
    scale = 1.0 * frame.shape[0] / scaled_frame.shape[0]
    img = gcv.utils.viz.cv_plot_bbox(frame.asnumpy(), bounding_boxes[0], scores[0], class_IDs[0], class_names=net.classes, scale=scale)
    gcv.utils.viz.cv_plot_image(img)
    cv2.waitKey(1)

cap.release()
cv2.destroyAllWindows()