hkchengrex/XMem
A PyTorch implementation of XMem, a deep learning video object segmentation system using an Atkinson-Shiffrin memory model to track objects across very long videos.
Not currently ranked — collecting fresh signals.
star history
XMem addresses video object segmentation by framing it as a memory problem, using an Atkinson-Shiffrin-inspired architecture to maintain object identity across frames over extended video sequences. The system handles occlusion and long-term dependencies through multi-level memory stores. Implemented in PyTorch, it includes a GUI for interactive video segmentation and achieves approximately 20 FPS on GPU.
Frequently asked
- What is hkchengrex/XMem?
- A PyTorch implementation of XMem, a deep learning video object segmentation system using an Atkinson-Shiffrin memory model to track objects across very long videos.
- Is XMem open source?
- Yes — hkchengrex/XMem is open source, released under the MIT license.
- What language is XMem written in?
- hkchengrex/XMem is primarily written in Python.
- How popular is XMem?
- hkchengrex/XMem has 2k stars on GitHub.
- Where can I find XMem?
- hkchengrex/XMem is on GitHub at https://github.com/hkchengrex/XMem.