PraNet: A Parallel Reverse Attention Network for Polyp Segmentation Colonoscopy is an effective technique for detecting colorectal polyps, which are closely related to colorectal cancer. Accurate polyp segmentation from colonoscopy images is crucial for diagnosis and surgery. However, this task is challenging due to the diversity of polyp appearances and the blurred boundary between polyps and surrounding mucosa. To address these challenges, we propose PraNet, a parallel reverse attention network for accurate polyp segmentation. PraNet uses a parallel partial decoder (PPD) to aggregate high-level features and generate a global map as initial guidance. It also employs reverse attention (RA) modules to mine boundary cues and establish relationships between areas and boundary cues. The recurrent cooperation between areas and boundaries allows PraNet to calibrate misaligned predictions and improve segmentation accuracy. Quantitative and qualitative evaluations on five challenging datasets across six metrics show that PraNet significantly improves segmentation accuracy, with advantages in generalizability and real-time segmentation efficiency (~50fps). PraNet is a deep neural network that combines a parallel partial decoder with reverse attention modules for accurate polyp segmentation. The PPD aggregates high-level features to generate a global map, while the RA modules mine boundary cues and establish relationships between areas and boundary cues. The recurrent cooperation between areas and boundaries allows PraNet to calibrate misaligned predictions and improve segmentation accuracy. PraNet outperforms existing methods on five challenging datasets, with significant improvements in segmentation accuracy and real-time performance. The model is efficient, with a training time of 20 epochs (~0.5 hours) and inference speed of ~50fps for 352x352 input. PraNet is also flexible and can be adapted to other tasks, such as lung infection segmentation. The model's performance is validated through extensive experiments on multiple datasets, demonstrating its effectiveness in polyp segmentation.PraNet: A Parallel Reverse Attention Network for Polyp Segmentation Colonoscopy is an effective technique for detecting colorectal polyps, which are closely related to colorectal cancer. Accurate polyp segmentation from colonoscopy images is crucial for diagnosis and surgery. However, this task is challenging due to the diversity of polyp appearances and the blurred boundary between polyps and surrounding mucosa. To address these challenges, we propose PraNet, a parallel reverse attention network for accurate polyp segmentation. PraNet uses a parallel partial decoder (PPD) to aggregate high-level features and generate a global map as initial guidance. It also employs reverse attention (RA) modules to mine boundary cues and establish relationships between areas and boundary cues. The recurrent cooperation between areas and boundaries allows PraNet to calibrate misaligned predictions and improve segmentation accuracy. Quantitative and qualitative evaluations on five challenging datasets across six metrics show that PraNet significantly improves segmentation accuracy, with advantages in generalizability and real-time segmentation efficiency (~50fps). PraNet is a deep neural network that combines a parallel partial decoder with reverse attention modules for accurate polyp segmentation. The PPD aggregates high-level features to generate a global map, while the RA modules mine boundary cues and establish relationships between areas and boundary cues. The recurrent cooperation between areas and boundaries allows PraNet to calibrate misaligned predictions and improve segmentation accuracy. PraNet outperforms existing methods on five challenging datasets, with significant improvements in segmentation accuracy and real-time performance. The model is efficient, with a training time of 20 epochs (~0.5 hours) and inference speed of ~50fps for 352x352 input. PraNet is also flexible and can be adapted to other tasks, such as lung infection segmentation. The model's performance is validated through extensive experiments on multiple datasets, demonstrating its effectiveness in polyp segmentation.