File size: 1,584 Bytes
133b124
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
from pathlib import Path
import base64, json, io
import cv2
from PIL import Image, ImageDraw

root=Path(__file__).resolve().parents[1]
registration=json.loads((root/'.local/grounded/registration-b86473ea-6c47-4755-bf19-8d7cbbd78804.json').read_text(encoding='utf-8'))
video=root.parent/'视频素材2/测试片段/01_典型事件/室内_smartbench_0188_幼犬翻越围栏.mp4'
binding=registration['binding'].copy();binding.pop('memoryId',None)
capture=cv2.VideoCapture(str(video))
def data(im):
    output=io.BytesIO();im.save(output,format='JPEG',quality=87)
    return 'data:image/jpeg;base64,'+base64.b64encode(output.getvalue()).decode()
def frame(t):
    capture.set(cv2.CAP_PROP_POS_MSEC,t*1000);ok,pixels=capture.read()
    if not ok:raise RuntimeError('Unable to decode selected video frame')
    im=Image.fromarray(cv2.cvtColor(pixels,cv2.COLOR_BGR2RGB));im.thumbnail((640,640))
    raw=data(im);marked=im.copy();draw=ImageDraw.Draw(marked);b=binding['regionBox'];w,h=marked.size
    draw.rectangle((round(b['x']*w),round(b['y']*h),min(w-1,round((b['x']+b['w'])*w)),min(h-1,round((b['y']+b['h'])*h))),outline='#34d399',width=3)
    return {'image':raw,'annotated':data(marked),'mediaTime':t,'frameId':'replay-'+str(t),'sourceId':'recorded-test-camera'}
result={'request':registration['request'],'binding':binding,'frames':[frame(t) for t in (0.1,1.1,2.1)],'observation':frame(3.1)}
capture.release();(root/'.local/app-api-fixture.json').write_text(json.dumps(result,ensure_ascii=False),encoding='utf-8')
print('Prepared recorded-video frames for real-model API acceptance')