vrw 0.1.0

Library for simple video reading and writing

vrw: A Simple Library for Video Reading and Writing

vrw is a lightweight Python library for reading and writing video files with slicing support. It provides a simple interface for handling video frames using NumPy arrays and supports multiple backends like OpenCV and PyAV.

---

Installation

To install vrw, use the following command:

git clone https://github.com/tkclam/vrw
cd vrw
pip install .

---

Usage

Reading Videos

The VideoReader class allows you to read video files and access frames with slicing and indexing.

from vrw import VideoReader

# Open a video file
vr = VideoReader("video.mp4")

# Access video properties
print(vr.shape)  # (N, H, W, C): N frames, H height, W width, C channels
print(vr.fps)    # Frames per second
print(vr.dtype)  # Data type of the frames

# Access frames using slicing
print(vr[:, 0].shape)         # (N, W, C): All frames, first row
print(vr[::2, :, -1].shape)   # (N // 2, H, C): Every second frame, last column
print(vr[..., 0].shape)       # (N, H, W): All frames, first channel
print(vr[[1, 3, 3, 7], :42].shape)  # (4, 42, W, C): Specific frames, first 42 rows

# Reading grayscale frames
vr_gray = VideoReader("video.mp4", to_gray=True)
print(vr_gray.shape)  # (N, H, W): Grayscale frames

Note that advanced indexing involving the 0th axis has not been implemented yet.

Writing Videos

The VideoWriter class allows you to write video files frame by frame.

Writing RGB Frames

from vrw import VideoWriter
import numpy as np

# Write RGB frames to a video file
with VideoWriter("video.mp4", fps=30) as vw:
    for i in range(256):
        im = np.full((256, 256, 3), i, dtype=np.uint8)
        im[..., 0], im[..., 1] = np.mgrid[:256, :256]
        vw.write(im)

Writing Grayscale Frames

# Write grayscale frames to a video file
with VideoWriter("video.mp4", fps=30) as vw:
    for i in range(256):
        vw.write(np.full((256, 256), i, dtype=np.uint8))

Using the PyAV Backend

The PyAV backend allows you to use advanced codecs (e.g., libx264) for video writing. Ensure that FFmpeg is installed and built with the required codecs.

# Use the PyAV backend with a specific codec
with VideoWriter("video.mp4", fps=30, backend="pyav", codec="libx264") as vw:
    for i in range(256):
        im = np.full((256, 256, 3), i, dtype=np.uint8)
        im[..., 0], im[..., 1] = np.mgrid[:256, :256]
        vw.write(im)

---

Backends

vrw supports the following backends

• OpenCV: Default backend for reading and writing videos.
• PyAV: Alternative backend for advanced codec support (requires FFmpeg built with the necessary codecs).

To specify a backend, use the backend parameter (e.g., backend="cv2" or backend="pyav") when creating a VideoReader or VideoWriter instance.