我想從視頻中獲取每一幀作為圖像。背景如下。我寫(xiě)了一個(gè)能夠識(shí)別手勢(shì)的神經(jīng)網(wǎng)絡(luò)?，F(xiàn)在我想開(kāi)始一個(gè)視頻流，其中流的每個(gè)圖像/幀都通過(guò)神經(jīng)網(wǎng)絡(luò)。為了使其適合我的神經(jīng)網(wǎng)絡(luò)，我想渲染每一幀并將圖像縮小到 28*28 像素。最后它看起來(lái)應(yīng)該類似于：https ://www.youtube.com/watch?v= JfSao30fMxY 我已經(jīng)通過(guò)網(wǎng)絡(luò)搜索并發(fā)現(xiàn)我可以使用 cv2.VideoCapture 來(lái)獲取流。但是我怎樣才能挑選幀的每個(gè)圖像，渲染它并將結(jié)果打印回屏幕上。到目前為止，我的代碼看起來(lái)像這樣：import numpy as npimport cv2cap = cv2.VideoCapture(0)# Todo: each Frame/Image from the video should be saved as a variable and open imageToLabel()# Todo: before the image is handed to the method, it needs to be translated into a 28*28 np Array# Todo: the returned Label should be printed onto the video (otherwise it can be )i = 0while (True):    # Capture frame-by-frame    # Load model once and pass it as an parameter    ret, frame = cap.read()    i += 1    image = cv2.imwrite('database/{index}.png'.format(index=i), frame)    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2BGRAY)    cv2.imshow('frame', frame)    if cv2.waitKey(1) & 0xFF == ord('q'):        break# When everything done, release the capturecap.release()cv2.destroyAllWindows()def imageToLabel(imgArr, checkpointLoad):    new_model = tf.keras.models.load_model(checkpointLoad)    imgArrNew = imgArr.reshape(1, 28, 28, 1) / 255    prediction = new_model.predict(imgArrNew)    label = np.argmax(prediction)    return label