def predict(hand):
img = cv2.resize(hand, (50, 50))
img = np.array(img)
img = img.reshape((1, 50, 50, 1))
img = img / 255.0
res = model.predict(img)

while True:
ret, frame = vc.read()
frame = cv2.flip(frame, 1)

cv2.rectangle(frame, (100, 100), (300, 300), (0, 255, 255), 2)
cv2.imshow("Recording", frame)

roi = frame[100:300, 100:300]

roi = cv2.cvtColor(roi, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(roi, (11, 11), 0)
blur = cv2.medianBlur(blur, 15)
thresh = cv2.threshold(blur, 210, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
thresh = cv2.bitwise_not(thresh)
cv2.imshow("hand", thresh)

count_frames += 1
print('Frame: ', count_frames)

if count_frames == 100:
    old_pred_text = pred_text
    #predict 100th frame
    pred_text = predict(thresh)

    if old_pred_text == pred_text:
        count_frames += 1
    else:
        count_frames = 0

    if count_frames > 20 and pred_text != "":
        tot_string += pred_text + " "
        count_frames = 0

k = cv2.waitKey(5) & 0xFF
if k == 27:
    break