Question

我正在使用 cv2.dnn 和 frozen_inference_graph.pb 在opencv中进行手势检测。代码工作正常，但问题是它真的很慢。

所以我想通过以下方式在gpu上运行代码：

cvNet = cv2.dnn.readNetFromTensorflow('model/frozen_inference_graph.pb', 'model/graph.pbtxt')
cvNet.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
cvNet.setPreferableTarget(cv2.dnn.DNN_TARGET_OPENCL_FP16)

从https://support.apple.com/en-us/HT202823中，我可以看到笔记本电脑中已经内置了opencl。我只是一个学生，所以我买不起NVIDIA GPU。

所有代码

import tensorflow as tf
import numpy as np
import time
import cv2 

cvNet = cv2.dnn.readNetFromTensorflow('model/frozen_inference_graph.pb', 'model/graph.pbtxt')
cvNet.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
cvNet.setPreferableTarget(cv2.dnn.DNN_TARGET_OPENCL_FP16)

capture = cv2.VideoCapture(0)

while True:
    # capture frame by frame
    ret, frame = capture.read()
    frame = cv2.flip(frame, 1)   

    rows = frame.shape[0]
    cols = frame.shape[1]
    cvNet.setInput(cv2.dnn.blobFromImage(frame, size=(300, 300), swapRB=True, crop=False))
    start = time.time()
    cvOut = cvNet.forward()

    end = time.time()
    print(end-start)
    for detection in cvOut[0, 0, :, :]:
        score = float(detection[2])
        if score > 0.3:
            left = detection[3] * cols
            top = detection[4] * rows
            right = detection[5] * cols
            bottom = detection[6] * rows
            cv2.rectangle(frame, (int(left), int(top)), (int(right),
                                                    int(bottom)), (23, 230, 210), thickness=2)

    cv2.imshow('img', frame)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cv2.release()
cv2.destroyAllWindows()

运行代码时，它使用了70％的cpu，但使用了0％的gpu。无论如何，我可以加快视频速度以获得更多fps吗？

谢谢。

无法在python的gpu上运行opencv dnn

0 个答案: