A Novel Model for Intelligent Pull-Ups Test Based on Key Point Estimation of Human Body and Equipment

Figure 4

The architecture of the proposed PEPoseNet. (a) The backbone being made of two functional network, i.e., heatmap network and key point network. (b) The structure of block 1, (c) the structure of block 2, and (d) the structure of block 3. The channels denoted by dotted arrow refer to that they have forward propagation only and have no gradient backward propagation. The “Conv” refers to standard convolution layer, while the depthwiseConv refers to depth-wise separable convolution [32–34].

(a)

(b)

(c)

(d)