[Paper Review] Feature Fusion(Alignment) Networks
1.[2017 CVPR] Feature Pyramid Network for Object Detection
paper : Feature Pyramid Networks for Object Detectionauthor : Tsung-Yi Lin, Piotr Dollár, Ross Girshick, Kaiming He, Bharath Hariharan, Serge Belongie
2.[2018 CVPR] Path Aggregation Network for Instance Segmentation
이 paper에서는, Path Aggregation Network (PANet)을 제안한다.구체적으로, bottom-up path augmentation을 통해 lower layeers의 accurate localization signals로 the entire fea
3.[2019 CVPR] NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection
현재 object detection을 위한 SOTA conv architectures는 manually designed되었다.우리는 OD를 위한 a better architecture of FPN을 학습하는 것을 목표로 한다.우리는 Neural Architecture
4.[2019 arXiv] Learning Spatial Fusion for Single-Shot Object Detection
pyramidal feature representation은 OD에서 common practice이다.(문제 제기)하지만, the inconsistency across different feature scales is a primary limitation for the
5.[2019 IEEE Access] Aggregated Residual Dilation-Based Feature Pyramid Network for Object Detection
multi-scale object detection을 위해서,top-down features and lateral features의 naive combination으로 생성되는hierarchical feature pyramids를 채택하는 것이 일반적이다.(문제 제기)
6.[2020 CVPR] EfficientDet에서 소개된 BiFPN
https://velog.io/@hseop/Simple-Review-EfficientDet-Scalable-and-Efficient-Object-Detection#3-bifpn
7.[2020 CVPR] AugFPN: Improving Multi-scale Feature Learning for Object Detection
Paper Info. Abstract
8.[중단] [2020 ECCV] Feature Pyramid Transformer
Paper Info. https://www.ecva.net/papers/eccv2020/papersECCV/papers/123730324.pdf Abstract
9.[2021 IEEE Access] Adaptive Feature Pyramid Networks for Object Detection
(문제점)Ghiasi et al.에 의해 제안된 feature pyramid network는 a simple fusion method를 채택했고, 이는 the fusion feature context를 고려하지 않아 실패했다.게다가, tranditional upsamp
10.[2021 ICCV] FaPN: Feature-aligned Pyramid Network for Dense Image Prediction
https://openaccess.thecvf.com/content/ICCV2021/papers/Huang_FaPN_Feature-Aligned_Pyramid_Network_for_Dense_Image_Prediction_ICCV_2021_paper.pdfmo
11.[2021 CVPR] A2-FPN: Attention Aggregation based Feature Pyramid Network for Instance Segmentation
https://openaccess.thecvf.com/content/CVPR2021/papers/Hu_A2-FPN_Attention_Aggregation_Based_Feature_Pyramid_Network_for_Instance_Segmentation_CVP
12.[2022 IEEE Access] CATFPN: Adaptive Feature Pyramid With Scale-Wise Concatenation and Self-Attention
heuristic feature fusion strategies에 기반한 FPN은 may be suboptimal이다.이 논문에서, 우리는 a novel FPN named CATFPN that consists of Scale-Wise Feature Concatenati
13.[2024] Slim-neck by GSConv: A lightweight-design for real-time detector architectures
Li, H., Li, J., Wei, H. et al. Slim-neck by GSConv: a lightweight-design for real-time detector architectures. J Real-Time Image Proc 21, 62 (2024). h
14.[2024 ECCV] (SlimNeckV2) Rethinking Features-Fused-Pyramid-Neck for Object Detection
Info. Abstract Multi-head detectors는 multi-scale detection을 위해 a features-fused-pyramid-neck을 사용하며, 이는 널리 채택되고 있다. 그러나, 이 approach는 서로 다른 hierarchi
15.[2021 WACV] Effective Fusion Factor in FPN for Tiny Object Detection
https://openaccess.thecvf.com/content/WACV2021/papers/Gong_Effective_Fusion_Factor_in_FPN_for_Tiny_Object_Detection_WACV_2021_paper.pdfFPN-based
16.[Simple Review] [2022 Applied Intelligence] Attention-based fusion factor in FPN for object detection
배경: 최근, 대부분의 detectors들은 다양한 크기의 objects를 detect하기 위해 feature pyramid를 활용함.문제 제기: 하지만 existing FPN-based feature extraction networks는 effective semant