Celux

Latest version: v0.6.0

Safety actively analyzes 714668 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 4 of 6

0.4.2

- **Focus on `VideoReader`**:
- Removed `VideoWriter` to streamline the library and enhance focus on reading capabilities.
- Fixed call method of `VideoReader`, now properly seeks frames.

- **New Features**:
- Added `__getitem__` method to `VideoReader` for easier access to properties, allowing users to retrieve metadata using dictionary-like syntax (e.g., `reader['width']`).
- Expanded `VideoReader.get_properties()` to include new metadata properties:
- `codec`: **The name of the codec being used.**
- `bit_depth`: **The bit-depth of the video.**
- `has_audio`: **Indicates whether the video contains an audio track.**
- `audio_bitrate`: **Bitrate of the audio stream.**
- `audio_channels`: **Number of audio channels.**
- `audio_sample_rate`: **Sample rate of the audio stream.**
- `audio_codec`: **Codec used for audio.**
- `min_fps`: **Minimum frames per second of the video.**
- `max_fps`: **Maximum frames per second of the video.**
- `aspect_ratio`: **Aspect ratio of the video.**

- **New Converter Formats**:
- Completed the implementation of the following converters to support new video formats:
- YUV420P to RGB
- YUV420P10LE to RGB48
- BGR to RGB
- RGB to RGB
- P010LE to RGB48

- **Supported Codecs**:
- The following codecs can be worked with using the `VideoReader`, based on supported pixel formats:
- **H.264 (AVC)**: YUV420P, YUV420P10LE
- **H.265 (HEVC)**: YUV420P, YUV420P10LE
- **VP8/VP9**: YUV420P, YUV420P10LE
- **AV1**: YUV420P, YUV420P10LE
- **MPEG-2**: YUV420P
- **ProRes**: YUV420P, YUV422, YUV444
- **DNxHD/DNxHR**: YUV422, YUV444
- **DV (Digital Video)**: YUV420P
- **Uncompressed RGB**: RGB, BGR
- **P010LE**: P010LE

- **Testing Improvements**:
- Updated tests to ensure compatibility with various bit-depths and codec types.
- Added tests to verify the correct functionality of the new features and converters.

0.4.0

- Moved to `FFmpeg` static libraries!
- Startup times are improved. All libs that can be static, are static.
- Adjusted logging to flow a little bit better, not overcrowd console unless desired.
- Logging details more info on codecs. The Decoder selects the **BEST** codec for the video.
- Need to investigate if `NVDEC` is bottlenecked, or I've reached max performance capabilities.
- It is curious that cpu benches at `1859 fps` and gpu benches at `1809 fps`.

0.3.9

- **Pre-Release Update:**
- Prep for **0.4.0** release.
- **0.4.x** release will be characterized by new codec and pixel format support!
- Removed `d_type` and `buffer_size` arguments from `VideoReader` and `VideoWriter`.
- Output and Input tensors are now, by standard, `UINT8`, `HWC` format, [0,255].
- Standardized to `YUV420P` for now.
- Swapped custom `CUDA` kernels for `nppi`.
- various cleanup and small refactorings.

0.3.8

- **Pre-Release Update:**
- Removed Buffering from `VideoWriter`, resulting in **INSANE** performance gains.
- Fixed threading issue with `VideoWriter`, now properly utilizes available threads.
- Removed `sync` method from `VideoWriter`.
- Synchronization can be manually handled by the user or by letting the `VideoWriter` do so on destruction.
- Updated Benchmarks to reflect new version.

0.3.7

- **Pre-Release Update:**
- Fixed remaining issues with `VideoWriter` class.
- Both `cpu` and `cuda` arguments NOW work properly.
- Few Small bug fixes regarding synchronization and memory management.

0.3.6

- **Pre-Release Update:**
- Fixed `VideoWriter` class.
- Both `cpu` and `cuda` arguments now work properly.
- **Encoder Functionality:**
- Enabled encoder support for both CPU and CUDA backends.
- Users can now encode videos directly from PyTorch tensors.
- Update Github Actions, add tests.

Page 4 of 6

© 2025 Safety CLI Cybersecurity Inc. All Rights Reserved.