human/render/video_render.py

#encoding = utf8
import copy
import time
from queue import Empty

import cv2
import numpy as np

from .base_render import BaseRender


class VideoRender(BaseRender):
    def __init__(self, play_clock, context, human_render):
        super().__init__(play_clock, 0.02)
        self._context = context
        self._human_render = human_render

    def _run_step(self):
        try:
            frame, ps = self._queue.get(block=True, timeout=0.01)
            res_frame, idx, type_ = frame
            print('video render queue size', self._queue.qsize())
        except Empty:
            return

        if type_ == 0:
            combine_frame = self._context.frame_list_cycle[idx]
        else:
            print('get face', self._queue.qsize())
            bbox = self._context.coord_list_cycle[idx]
            combine_frame = copy.deepcopy(self._context.frame_list_cycle[idx])
            y1, y2, x1, x2 = bbox
            try:
                res_frame = cv2.resize(res_frame.astype(np.uint8), (x2 - x1, y2 - y1))
            except:
                print('resize error')
                return
            # combine_frame = get_image(ori_frame,res_frame,bbox)
            # t=time.perf_counter()
            combine_frame[y1:y2, x1:x2] = res_frame

        clock_time = self._play_clock.clock_time()
        time_difference = clock_time - ps

        print('video render:', ps, '  ', clock_time, '  ', time_difference)
        if time_difference < -0.01:  # 音频比视频快超过10ms
            sleep_time = abs(time_difference + 0.01)
            print("Video frame waiting to catch up with audio", sleep_time)
            if sleep_time > 0:
                time.sleep(sleep_time)  # 只在正值时调用 sleep
            return  # 继续等待

        elif time_difference < -0.01:  # 视频比音频快超过10ms
            print("Video frame dropped to catch up with audio")
            return  # 丢帧
        # if time_difference > self._play_clock.audio_diff_threshold:
        #     # print('video is slow')
        #     return
        # elif time_difference < self._play_clock.audio_diff_threshold:

        image = combine_frame
        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        if self._human_render is not None:
            self._human_render.put_image(image)