Перебрать все кадры в видео питоне

Question

Я новичок, и я хочу прочитать видео и определить лица, разрезав его на кадры, но это не сработало, и я не понял, в чем проблема. не могли бы вы помочь мне исправить это, пожалуйста. NB :: Я работаю с Google Colaboraty и обнаруживаю лица с помощью библиотеки face_recognition, там мой код:

import face_recognition
import cv2
from google.colab.patches import cv2_imshow
input_video = cv2.VideoCapture('/content/My Drive/video-3.mp4')
# Metadata from the input video
frames_per_second = int(input_video.get(cv2.CAP_PROP_FPS))
frame_width = int(input_video.get(cv2.CAP_PROP_FRAME_WIDTH))
frame_height = int(input_video.get(cv2.CAP_PROP_FRAME_HEIGHT))

print('Metadata from input video:',
      '\nFrames per second:', frames_per_second,
      '\nFrame width:', frame_width, 
      '\nFrame height:', frame_height)
codec = cv2.VideoWriter.fourcc(*'XVID')
video_writer = cv2.VideoWriter('output_video.mp4', 
                               codec, 
                               frames_per_second, 
                               (frame_width, frame_height))
# An array to hold the locations of faces that are detected on individual frames
face_locations = []

# A counter to keep track of the number of frames processed
count = 1

# Loop through all the frames in the video
while (count != no_of_frames):
  # Read the video to retrieve individual frames. 'frame' will reference the inidivdual frames read from the video.
  ret, frame = input_video.read()

  # Check the 'ret' (return value) to see if we have read all the frames in the video to exit the loop
  if not ret:
    print('Processed all frames')
    break

  # Convert the image (frame) to RGB format as by default Open CV uses BGR format. 
  # This conversion is done as face_recognition and other libraries usually use RGB format.
  rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

  # Get the coordinates in the image where a face is detected. Use the model 'cnn' after greater accuracy.
  face_locations = face_recognition.face_locations(rgb_frame, model='cnn')

  # Loop through the face locations array and draw a rectangle around each face that is detected in the frame
  for top, right, bottom, left in face_locations:
    cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2)

  # Write the frame to the output vide0
  video_writer.write(frame)

  # Print for every 50 frames processed
  if(count % 50 == 0):
    print('Processed', count, 'frames')

  count += 1

# Release to close all the resources that we have opened for reading and writing video
input_video.release()
video_writer.release() 

cv2.destroyAllWindows()

Результат: в строке 7: имя 'no_of_frames' не определено

---

Ее это полный код

import face_recognition
import cv2
from google.colab.patches import cv2_imshow
input_video = cv2.VideoCapture('/content/My Drive/video-3.mp4')
# Metadata from the input video
frames_per_second = int(input_video.get(cv2.CAP_PROP_FPS))
frame_width = int(input_video.get(cv2.CAP_PROP_FRAME_WIDTH))
frame_height = int(input_video.get(cv2.CAP_PROP_FRAME_HEIGHT))

print('Metadata from input video:',
      '\nFrames per second:', frames_per_second,
      '\nFrame width:', frame_width, 
      '\nFrame height:', frame_height)
codec = cv2.VideoWriter.fourcc(*'XVID')
video_writer = cv2.VideoWriter('output_video.mp4', 
                               codec, 
                               frames_per_second, 
                               (frame_width, frame_height))
# An array to hold the locations of faces that are detected on individual frames
face_locations = []

# A counter to keep track of the number of frames processed
count = 1

# Loop through all the frames in the video
while (count != no_of_frames):
  # Read the video to retrieve individual frames. 'frame' will reference the inidivdual frames read from the video.
  ret, frame = input_video.read()

  # Check the 'ret' (return value) to see if we have read all the frames in the video to exit the loop
  if not ret:
    print('Processed all frames')
    break

  # Convert the image (frame) to RGB format as by default Open CV uses BGR format. 
  # This conversion is done as face_recognition and other libraries usually use RGB format.
  rgb_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)

  # Get the coordinates in the image where a face is detected. Use the model 'cnn' after greater accuracy.
  face_locations = face_recognition.face_locations(rgb_frame, model='cnn')

  # Loop through the face locations array and draw a rectangle around each face that is detected in the frame
  for top, right, bottom, left in face_locations:
    cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2)

  # Write the frame to the output vide0
  video_writer.write(frame)

  # Print for every 50 frames processed
  if(count % 50 == 0):
    print('Processed', count, 'frames')

  count += 1

# Release to close all the resources that we have opened for reading and writing video
input_video.release()
video_writer.release()
cv2.destroyAllWindows()

Answer 1

Вы не объявили переменную no_of_frames. Просто как тот. Вот пример кода, который выполняет ту же задачу.

capture = cv2.VideoCapture(0)
while 1:
    _, image = capture.read()

    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

    faces = face_cascade.detectMultiScale(gray, 1.3, 5)

    for (x,y,w,h) in faces:
        cv2.circle(image,(x+int(w/2),y+int(h/2)),(int(h)),(255,0,0),4)
        roi_gray = gray[y:y+h, x:x+w]
        roi_color = image[y:y+h, x:x+w]

        eyes = eye_cascade.detectMultiScale(roi_gray)

        for (ex,ey,ew,eh) in eyes:
            cv2.circle(roi_color, (ex + int(ew/2), ey + int(eh/2)), (int(eh/2)), (0,0,255), 2)

    cv2.imshow('Image', image)

    k = cv2.waitKey(30) & 0xff

    if k == 27:
        break

# Close the window
capture.release()

# De-allocate any associated memory usage
cv2.destroyAllWindows()