【更新0.16.1正式版】万字经验，一文带你入门frigate AI监控

a784174467

gasment 发表于 2025-9-3 15:46
我推测是zone区域的视角问题，黄色zone区域视角可能过于垂直，能捉到的猫特征减少，导致置信度下降，不足 ...

哦哦，好的，明白了！感谢回复！我试试看！

tongshome

感谢分享

a784174467

请问这张图片里，猫只是红色框，没有label和置信度，是不是需要调整yaml呀？这是我现在的yaml：
cameras:
Cat_camera: # <--- This is the relevant camera
enabled: true
ffmpeg:
inputs:
- path: rtsp://<ip_address>/stream1
roles:
- detect
- path: rtsp://<ip_address>/stream2
roles:
- record
motion:
mask: 0.377,0.052,0.504,0.051,0.509,0.102,0.384,0.107
objects:
filters:
person: {}
cat: {}
zones:
cat_toilet: # <--- This is the relevant zone
coordinates:
0.398,0.724,0.383,0.683,0.343,0.564,0.282,0.511,0.128,0.656,0.208,1,0.451,0.894,0.42,0.808
loitering_time: 0
objects: cat
inertia: 3
version: 0.16-0

detect:
fps: 5
stationary:
interval: 20
threshold: 50
enabled: true
objects:
track:
- person
- cat
filters:
cat:
min_area: 5000
min_score: 0.45
threshold: 0.55

detectors:
ov1:
type: openvino
device: CPU
num_threads: 8
model:
model_type: yolonas
width: 320
height: 320
input_tensor: nchw
input_pixel_format: bgr
path: /config/yolo_nas_s.onnx
labelmap_path: /labelmap/coco-80.txt

fengqs

感谢大佬,头疼了好几天.

li23334

感谢分享

goodhelp

优秀，真的用心

623305170

真不错，加油

fdg106

感谢分享

labelzhou

支持下, 看看打包的模型地址

wanwangzi

谢谢大佬