GOATmessi8/ASFF
A PyTorch implementation of Adaptively Spatial Feature Fusion (ASFF) for single-shot object detection, built on YOLOv3.
Not currently ranked — collecting fresh signals.
star history
This work introduces ASFF, a feature fusion strategy that learns to filter conflicting information across different spatial scales to improve detection accuracy. The implementation modifies YOLOv3’s architecture with an attention-based fusion mechanism and supports multiple backbone options including MobileNetV2. It achieves improved mAP on COCO benchmarks while maintaining reasonable inference speeds compared to baseline YOLOv3.
Frequently asked
- What is GOATmessi8/ASFF?
- A PyTorch implementation of Adaptively Spatial Feature Fusion (ASFF) for single-shot object detection, built on YOLOv3.
- Is ASFF open source?
- Yes — GOATmessi8/ASFF is open source, released under the GPL-3.0 license.
- What language is ASFF written in?
- GOATmessi8/ASFF is primarily written in Python.
- How popular is ASFF?
- GOATmessi8/ASFF has 1.1k stars on GitHub.
- Where can I find ASFF?
- GOATmessi8/ASFF is on GitHub at https://github.com/GOATmessi8/ASFF.