delayless video detector

[jackyetz]

As the video frame detector causes playback delay, I modified the "detect_video.py" for delayless video detection. Two parameters introduced, where "buffersize" reading consecutive frames ahead, "delay" processing a future frame within the buffered frames.

`import time from absl import app, flags, logging from absl.flags import FLAGS import cv2 import tensorflow as tf from yolov3_tf2.models import ( YoloV3, YoloV3Tiny ) from yolov3_tf2.dataset import transform_images from yolov3_tf2.utils import draw_outputs import threading import queue queue_img = queue.Queue() buffersize = 20

flags.DEFINE_string('classes', './data/coco.names', 'path to classes file') flags.DEFINE_string('weights', './checkpoints/yolov3.tf', 'path to weights file') flags.DEFINE_boolean('tiny', False, 'yolov3 or yolov3-tiny') flags.DEFINE_integer('size', 416, 'resize images to') flags.DEFINE_string('video', './data/video.mp4', 'path to video file or number for webcam)') flags.DEFINE_integer('timeF', 3, '视频帧计数间隔频率ms)') flags.DEFINE_string('output', None, 'path to output video') flags.DEFINE_string('output_format', 'XVID', 'codec used in VideoWriter when saving video to file') flags.DEFINE_integer('num_classes', 80, 'number of classes in the model') flags.DEFINE_integer('delay', 10, 'delay frames')

def main(_argv): physical_devices = tf.config.experimental.list_physical_devices('GPU') for physical_device in physical_devices: tf.config.experimental.set_memory_growth(physical_device, True)

if FLAGS.tiny:
    yolo = YoloV3Tiny(classes=FLAGS.num_classes)
else:
    yolo = YoloV3(classes=FLAGS.num_classes)

yolo.load_weights(FLAGS.weights)
logging.info('weights loaded')

class_names = [c.strip() for c in open(FLAGS.classes).readlines()]
logging.info('classes loaded')

times = []

try:
    vid = cv2.VideoCapture(int(FLAGS.video))
except:
    vid = cv2.VideoCapture(FLAGS.video)

out = None

if FLAGS.output:
    # by default VideoCapture returns float instead of int
    width = int(vid.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(vid.get(cv2.CAP_PROP_FRAME_HEIGHT))
    fps = int(vid.get(cv2.CAP_PROP_FPS))
    codec = cv2.VideoWriter_fourcc(*FLAGS.output_format)
    out = cv2.VideoWriter(FLAGS.output, codec, fps, (width, height))

boxes, scores, classes, nums, timeslap = None, None, None, None, None
threadrun = MyThread(yolo, queue_img, boxes, scores, classes, nums, timeslap)
threadrun.setDaemon(True)
threadrun.start()
while True:
    imgs = []
    for _ in range(buffersize):
        ret, img = vid.read()
        if img is None: break
        imgs.append(img)
    if len(imgs)==0:
        logging.warning("Empty Frame")
        time.sleep(0.1)
        continue
    imgs_in = [cv2.cvtColor(img, cv2.COLOR_BGR2RGB) for img in imgs]
    imgs_in = [tf.expand_dims(img_in, 0) for img_in in imgs_in]
    imgs_in = [transform_images(img_in, FLAGS.size) for img_in in imgs_in]

    for index, img in enumerate(imgs):
        if queue_img.empty(): # 线程没有图片处理的时候，输送最新图片给线程
            queue_img.put(imgs_in[min(len(imgs_in)-1,index+FLAGS.delay)])
        while timeslap==None:
            time.sleep(0.02)
            boxes, scores, classes, nums, timeslap = threadrun.get_result()
        boxes, scores, classes, nums, timeslap = threadrun.get_result()
        times.append(timeslap)
        times = times[-20:]
        img = draw_outputs(img, (boxes, scores, classes, nums), class_names)
        img = cv2.putText(img, "Time: {:.2f}ms".format(sum(times) / len(times) * 1000), (0, 30),
                          cv2.FONT_HERSHEY_COMPLEX_SMALL, 1, (0, 0, 255), 2)
        if FLAGS.output:
            out.write(img)
        cv2.imshow('output', img)
        if cv2.waitKey(1) == ord('q'):
            break
        cv2.waitKey(30)

cv2.destroyAllWindows()
vid.release()

class MyThread(threading.Thread): def init(self, model, queue_img, boxes, scores, classes, nums, timeslap): threading.Thread.init(self) self.model = model self.queue = queue_img self.b,self.s,self.c,self.n,self.t=boxes, scores, classes, nums, timeslap def run(self): while True: try: t1 = time.time() self.boxes, self.scores, self.classes, self.nums = self.model.predict(self.queue.get()) if not self.queue.empty(): self.queue.get() t2 = time.time() self.timeslap = t2-t1 except Exception: pass # signals to queue job is done self.queue.task_done() def get_result(self): try: return self.boxes, self.scores, self.classes, self.nums, self.timeslap except Exception: return self.b,self.s,self.c,self.n,self.t

if name == 'main': try: app.run(main) except SystemExit: pass `