MilInst: Enhanced Instance Segmentation Framework for Military Camouflaged Targets Using Sparse Instance Activation

In this study, an improved end-to-end framework for instance segmentation of military camouflaged targets, referred to as MilInst, is proposed. The builds upon SparseInst method developed by Cheng et al . [27]. Several improvements are introduced enhance the model’s performance. First, Receptive Field Enhancement Module (RFEM) employed capture broader contextual information. Additionally, Feature Merging (FMM) utilized eliminate feature noise through implementation a matrix decomposition method. Furthermore, novel linear dynamic bipartite matching approach proposed, facilitating smooth transition from one-to-one one-to-many, and eventually achieving accurate matching. Experimental results demonstrate effectiveness MilInst algorithm. Comparisons with selected real-time baseline model reveal superior performance, highest mean Average Precision (mAP) index 84.8% on self-made dataset.