Image Semantic Segmentation Use Multiple-Threshold Probabilistic R-CNN with Feature Fusion

With continuous developments in deep learning, image semantic segmentation technology has also undergone great advancements and been widely used many fields with higher accuracy. This paper proposes an algorithm based on a neural network. Based the Mask Scoring R-CNN, this uses symmetrical feature pyramid network adds multiple-threshold architecture to improve sample screening precision. We employ probability model optimize mask branch of further accuracy for edges. In addition, we adjust loss function so that experimental effect can be optimized. The experiments reveal improves results.