使用vscode 的 debug 调试 使用openvino推理可能会出现问题.使用.pt推理没问题 #533
Description
最小复现代码
import cv2
from ultralytics import YOLOv10
# 读取并调整图像尺寸
img = cv2.imread("test.jpg")
img = cv2.resize(img, (640, 640))
ov_model = YOLOv10("best_openvino_model/")
result = ov_model.predict(source=img, imgsz=640, conf=0.7, save=False)
print(result[0].boxes.xywhn)
result = ov_model.predict(source=img, imgsz=640, conf=0.7, save=False)
print(result[0].boxes.xywhn)
pt_model = YOLOv10("best.pt")
result = pt_model.predict(source=img, imgsz=640, conf=0.7, save=False)
print(result[0].boxes.xywhn)
result = pt_model.predict(source=img, imgsz=640, conf=0.7, save=False)
print(result[0].boxes.xywhn)在每个print的位置打断点,依次运行
可以观察到打印
WARNING ⚠️ Unable to automatically guess model task, assuming 'task=detect'. Explicitly define task for your model, i.e. 'task=detect', 'segment', 'classify','pose' or 'obb'.
Loading best_openvino_model for OpenVINO inference...
Using OpenVINO LATENCY mode for batch=1 inference...
0: 640x640 1 Conehead Zombie, 3 Zombies, 61.0ms
Speed: 8.6ms preprocess, 61.0ms inference, 1.0ms postprocess per image at shape (1, 3, 640, 640)
Backend qtagg is interactive backend. Turning interactive mode on.
tensor([[0.6655, 0.3269, 0.0894, 0.2034],
[0.2181, 0.6775, 0.1150, 0.1995],
[0.6413, 0.1660, 0.1052, 0.1714],
[0.4406, 0.8378, 0.1043, 0.1980]])
0: 640x640 (no detections), 33.0ms
Speed: 3.6ms preprocess, 33.0ms inference, 0.0ms postprocess per image at shape (1, 3, 640, 640)
tensor([], size=(0, 4))
0: 640x640 1 Conehead Zombie, 3 Zombies, 149.1ms
Speed: 3.0ms preprocess, 149.1ms inference, 1.0ms postprocess per image at shape (1, 3, 640, 640)
tensor([[0.6655, 0.3269, 0.0894, 0.2034],
[0.2181, 0.6775, 0.1150, 0.1995],
[0.6413, 0.1660, 0.1052, 0.1714],
[0.4406, 0.8378, 0.1043, 0.1980]])
0: 640x640 1 Conehead Zombie, 3 Zombies, 123.4ms
Speed: 2.0ms preprocess, 123.4ms inference, 1.0ms postprocess per image at shape (1, 3, 640, 640)
tensor([[0.6655, 0.3269, 0.0894, 0.2034],
[0.2181, 0.6775, 0.1150, 0.1995],
[0.6413, 0.1660, 0.1052, 0.1714],
[0.4406, 0.8378, 0.1043, 0.1980]])
即使用openvino推理的时候,第二次推理获取不到结果，而pt推理是正常的