Serial-parallel multi-scale feature fusion for anatomy-oriented hand joint detection

Accurate hand joints detection from images is a fundamental topic that essential for many applications in computer vision and human–computer interaction. This paper presents two-stage network single unmarked image by using serial-parallel multi-scale feature fusion. In stage I, the regions are located an encoder-decoder network, features of each detected region extracted shallow spatial representation module. The then fed into II, which consists serially connected extraction modules with similar structures, called “multi-scale fusion” (MSFF). An MSFF contains parallel branches, generate initial joint heatmaps. heatmaps mutually reinforced anatomic relationship between joints. accuracy shows proposed overperforms state-of-the-art methods on current datasets, 1) RHD, 2) HS, 3) MPII & NZSL, 4) DCD8-6000, [email protected] 0.94, 0.92, 0.84, 0.97. Meanwhile, one takes 24 37 ms to process, adequate supporting real-time applications.