Efficiently identify people in image in one stage without separate detection step
Efficiently identify people in image in one stage without separate detection step
Efficient Person Search: An Anchor-Free Approach
arXiv paper abstract https://arxiv.org/abs/2109.00211v1
arXiv PDF paper https://arxiv.org/pdf/2109.00211v1.pdf
GitHub https://github.com/daodaofr/AlignPS
Person search aims to simultaneously localize and identify a query person from realistic, uncropped images. … models typically add a re-id branch upon two-stage detectors like Faster R-CNN.
… but … introduces high computational overhead due to dense object anchors.
… First, we select an anchor-free detector (i.e., FCOS) as the prototype of our framework. … exhibits significantly higher efficiency
… propose an aligned feature aggregation module to generate more discriminative and robust feature embeddings.
… name our model as Feature-Aligned Person Search Network (AlignPS).
… augment AlignPS with an ROI-Align head, which significantly improves the robustness
… framework achieves state-of-the-art or competitive performance, while displaying higher efficiency. …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
