DoubleHigherNet: Coarse-to-Fine Precise Heatmap Bottom- Up Dynamic Pose Computer Intelligent Estimation

Accurate keypoint positioning is necessary for bottom-up multi-person pose estimation methods to handle scale variation and crowdedness. In this paper, we present DoubleHigherNet: a novel network learning scale-aware precise heatmap representation process using double high-resolution feature pyramids coarse-to-fine training. The two in DoubleHigherNet consists of 1/4 resolution higher-resolution (1/2) maps generated by attention fusion blocks transposed convolutions. Benefited the training strategy, muti-resoltion coarse-fine aggregation, proposed approach able predict keypoints more accurately so as perform better on difficult crowded scenes. DoubleHigherNet-w32 achieves competitive result CrowdPose-test, surpassing all top-down SOTA HigherHRNet-w32 (which possesses similar number params with DoubleHigherNet-w32).