search

jiang-du / BlazePose-tensorflow

A third-party Tensorflow Implementation for paper "BlazePose: On-device Real-time Body Pose tracking".
Apache License 2.0
113 stars 24 forks source link

Question about model architecture #2

Open ejmejm opened 4 years ago

ejmejm commented 4 years ago

Hello, I was looking through the implementation, and I was curious about how you made choices regarding what to use for the architecture. Looking at the original paper here, I saw conv filter sizes and channel numbers listed, but I don't see any listings of the number of blaze blocks to use per layer and such parameters listed.

I used this paper as a reference to understand the structure of BlazeBlocks, but is that what you based your implementation on? Or did you get your architecture parameters from a different paper, or did you just experiment with it yourself?

jiang-du commented 4 years ago

The original paper on arxiv did not mention the detailed architecture of the network. While, on the google AI blog, the authors released an implementation for mobile APP. My previous selection of the number of blaze blocks and channels was mainly based on the visualization of their model.