Recherche avancée

Sur d’autres sites (6070)

• Tensorflow Video Parsing : ValueError : Convert None with Unsupported Type of class 'NoneType

  17 février 2024, par John Siddarth

  I am trying to learn how to use classifications on video data. I try to run an example from Tensorflow.

  


  I installed related tools using this command :

  


  # The way this tutorial uses the `TimeDistributed` layer requires TF>=2.10
pip install -U "tensorflow>=2.10.0"

pip install remotezip tqdm opencv-python
pip install -q git+https://github.com/tensorflow/docs


  


  Download the video file by curl -O https://upload.wikimedia.org/wikipedia/commons/8/86/End_of_a_jam.ogv.

  


  The code :

  


  import tqdm
import random
import pathlib
import itertools
import collections

import os
import cv2
import numpy as np
import remotezip as rz

import tensorflow as tf

# Some modules to display an animation using imageio.
import imageio
from IPython import display
from urllib import request
from tensorflow_docs.vis import embed

def format_frames(frame, output_size):
  """
    Pad and resize an image from a video.

    Args:
      frame: Image that needs to resized and padded. 
      output_size: Pixel size of the output frame image.

    Return:
      Formatted frame with padding of specified output size.
  """
  frame = tf.image.convert_image_dtype(frame, tf.float32)
  frame = tf.image.resize_with_pad(frame, *output_size)
  return frame

def frames_from_video_file(video_path, n_frames, output_size = (224,224), frame_step = 15):
  """
    Creates frames from each video file present for each category.

    Args:
      video_path: File path to the video.
      n_frames: Number of frames to be created per video file.
      output_size: Pixel size of the output frame image.

    Return:
      An NumPy array of frames in the shape of (n_frames, height, width, channels).
  """
  # Read each video frame by frame
  result = []
  src = cv2.VideoCapture(str(video_path))  

  video_length = src.get(cv2.CAP_PROP_FRAME_COUNT)

  need_length = 1 + (n_frames - 1) * frame_step

  if need_length > video_length:
    start = 0
  else:
    max_start = video_length - need_length
    start = random.randint(0, max_start + 1)

  src.set(cv2.CAP_PROP_POS_FRAMES, start)
  # ret is a boolean indicating whether read was successful, frame is the image itself
  ret, frame = src.read()
  result.append(format_frames(frame, output_size))

  for _ in range(n_frames - 1):
    for _ in range(frame_step):
      ret, frame = src.read()
    if ret:
      frame = format_frames(frame, output_size)
      result.append(frame)
    else:
      result.append(np.zeros_like(result[0]))
  src.release()
  result = np.array(result)[..., [2, 1, 0]]

  return result

video_path = "End_of_a_jam.ogv"

sample_video = frames_from_video_file(video_path, n_frames = 10)
sample_video.shape


  


  Here are the summary of the troubleshooting I did :

  


  


  • Check File Path : I verified that the file path to the video was correct and accessible from the Python environment.

    


  


  • Verify File Permissions : I ensured that the video file had the necessary permissions to be read by the Python code.

    


  


  • Test with Absolute Path : I attempted to use an absolute file path to access the video file to eliminate any ambiguity in the file's location.

    


  


  • Check File Format and Encoding : I examined the video file to ensure it was in a supported format and encoded properly for reading by OpenCV.

    


  


  


  What could be a cause ?

  


• libavfilter/vf_dnn_detect : Use class confidence to filt boxes

  17 janvier 2024, par Wenbin Chen

  libavfilter/vf_dnn_detect : Use class confidence to filt boxes
  Use class confidence instead of box_score to filt boxes, which is more
  
  accurate. Class confidence is obtained by multiplying class probability
  
  distribution and box_score.
  Signed-off-by : Wenbin Chen <wenbin.chen@intel.com>
  
  Reviewed-by : Guo Yejun <yejun.guo@intel.com>
  

  • [DH] libavfilter/vf_dnn_detect.c

• avformat/options : add missing disposition flag to AVStream class options

  25 octobre 2023, par James Almer

  avformat/options : add missing disposition flag to AVStream class options
  Signed-off-by : James Almer <jamrial@gmail.com>
  

  • [DH] libavformat/options.c
  • [DH] libavformat/version.h