#259 · Primary category： Computer Vision
decord
An efficient video loader for deep learning with smart shuffling that's super easy to digest
Project last updated：07/17/24
GitHub Stars
2.5K
Forks
233
Contributors
12
License
Apache-2.0
Why we included this project
Video-model training often spends most of its time just grabbing frames out of files, and Decord exists to cut that overhead. It wraps hardware-accelerated decoders (FFmpeg/LibAV, NVIDIA, Intel) behind a simple Python slicing API, so you can pull any frame or short clip without re-decoding the whole video. Random access is the core strength, and its built-in shuffling keeps data loading as smooth as an image loader does for batches. If you are feeding video clips into PyTorch, MXNet, or TensorFlow, the automatic array conversion and synchronized audio-video decoding will save you real integration work. Just know the pip package is CPU-only; using the GPU codecs means building from source, so plan that into your setup.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
opencv
Open Source Computer Vision Library
RuView
π RuView turns commodity WiFi signals into real-time spatial intelligence, vital sign monitoring, and presence detection — all without a single pixel of video.
PaddleOCR
Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.
MinerU
Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
tesseract
Tesseract Open Source OCR Engine (main repository)