Fast and Lightweight Human Pose Estimation

Although achieving significant improvement on pose estimation, the major drawback is that most top-performing methods tend to adopt complex architecture and spend large computational cost achieve higher performance. Due edge device’s limited resources, its are hard maintain fast inference speed in practice. To address this issue, we proposed lightweight human estimation method high performance bear less cost. Especially, consists of two parts, i.e., network (FLPN) for a novel bottleneck block reducing cost, which can integrate simple into an efficient accurate estimation. In terms block, introduce structural similarity measurement (SSIM) refine appropriate ratio intrinsic feature maps reduce model size. Furthermore, attention mechanism also adopted our modeling contextual information. We demonstrate with extensive experiments standard benchmark datasets by comparing state-of-the-art methods. On COCO keypoint detection dataset, attains similar accuracy these methods, but more than 7 times ours.