pengzhiliang/Conformer
A hybrid CNN-Transformer architecture published at ICCV21 for visual recognition tasks.
Not currently ranked — collecting fresh signals.
star history
Conformer combines convolutional neural networks and visual transformers to leverage both local feature extraction and global representation capture. The model uses a Feature Coupling Unit (FCU) to fuse local features with global representations interactively across different resolutions, maintaining a concurrent dual-branch structure. It serves as a general-purpose backbone for image classification, object detection, and instance segmentation tasks.
Frequently asked
- What is pengzhiliang/Conformer?
- A hybrid CNN-Transformer architecture published at ICCV21 for visual recognition tasks.
- Is Conformer open source?
- Yes — pengzhiliang/Conformer is open source, released under the Apache-2.0 license.
- What language is Conformer written in?
- pengzhiliang/Conformer is primarily written in Jupyter Notebook.
- How popular is Conformer?
- pengzhiliang/Conformer has 600 stars on GitHub.
- Where can I find Conformer?
- pengzhiliang/Conformer is on GitHub at https://github.com/pengzhiliang/Conformer.