/avi-r

AVI-R Package (formerly DIVA IO): A robust reader for AVI video files

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

AVI-R Package

PyPI version Downloads Publish to PyPI (formerly DIVA IO)

Author: Lijun Yu

Email: lijun@lj-y.com

A robust reader for AVI video files. Originally designed for the MEVA dataset to replace OpenCV's cv2.VideoCapture in the DIVA project.

Installation

pip install avi-r

Usage

A robust video loader that deals with missing frames in AVI files. In AVI-R, missing frames are automatically filled with the previous available frame, or the next available frame for the beginning of a video. This ensures you are getting the correct frame ids and valid frame contents all the time.

In comparison, OpenCV's cv2.VideoCapture would skip missing frames without warning, leading to wrong frame ids. Pims would return empty frames also without warning. And decord would crash.

Iterator Interface

from avi_r import AVIReader
video = AVIReader(video_path) # or AVIReader(video_name, parent_dir)
for frame in video:
    # frame is a avi_r.frame.Frame object
    image = frame.numpy()
    # image is an uint8 array in a shape of (height, width, channel[BGR])
    # ... Do something with the image
video.close() # Release internal buffers

Replace cv2.VideoCapture

To replace the cv2.VideoCapture objects in legacy codes, simply change from

import cv2
cap = cv2.VideoCapture(video_path)

to

from avi_r import AVIReader
cap = AVIReader(video_path)

AVIReader.read follows the schema of cv2.VideoCapture.read but automatically inserts the missing frames while reading the video. AVIReader.release also follows cv2.VideoCapture.release.

Random Access

Random access of a frame requires decoding from the nearest key frame (approximately every 60 frames for MEVA). Averagely, this introduces a constant overhead of 0.1 seconds. However, when the nearest key frame is missing, AVIReader will try to search a valid one till the beginning.

start_frame_id = 1500
length = 100
video.seek(start_frame_id)
for frame in video.get_iter(length):
    image = frame.numpy()
    # ... Do something with the image
video.close()

Video Properties

video.num_frames # cap.get(cv2.CAP_PROP_FRAME_COUNT)
video.frame_rate # cap.get(cv2.CAP_PROP_FPS)
video.height # cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
video.width # cap.get(cv2.CAP_PROP_FRAME_WIDTH)

Other Interfaces

For other usages, please see the comments in reader.py.

Speed

See speed.md.

Version History

See version.md.