The paper introduces the Path Aggregation Network (PANet) for instance segmentation, aiming to enhance information flow in proposal-based frameworks. PANet improves the feature hierarchy by augmenting bottom-up paths to propagate accurate localization signals from lower layers to the topmost feature, reducing the information path length. It also introduces adaptive feature pooling to link feature grids and all feature levels, allowing useful information from each level to be directly used by proposal subnetworks. Additionally, a complementary branch captures different views for each proposal to further improve mask prediction. These improvements are simple to implement with minimal computational overhead. PANet achieves state-of-the-art performance on the COCO 2017 Challenge Instance Segmentation task and second place in the Object Detection task without large-batch training. It also performs well on MVD and Cityscapes datasets. The code is available at <https://github.com/ShuLiu1993/PANet>.The paper introduces the Path Aggregation Network (PANet) for instance segmentation, aiming to enhance information flow in proposal-based frameworks. PANet improves the feature hierarchy by augmenting bottom-up paths to propagate accurate localization signals from lower layers to the topmost feature, reducing the information path length. It also introduces adaptive feature pooling to link feature grids and all feature levels, allowing useful information from each level to be directly used by proposal subnetworks. Additionally, a complementary branch captures different views for each proposal to further improve mask prediction. These improvements are simple to implement with minimal computational overhead. PANet achieves state-of-the-art performance on the COCO 2017 Challenge Instance Segmentation task and second place in the Object Detection task without large-batch training. It also performs well on MVD and Cityscapes datasets. The code is available at <https://github.com/ShuLiu1993/PANet>.