Mobilenet-V2 variants have superior performances over previous models(v1).
I have tried to train new models with the new backbone architecture.
Training is on-going.
https://github.com/ildoonet/tf-pose-estimation/issues/369
Need to check the performance with the proper preprocessing.
https://github.com/ildoonet/tf-pose-estimation/blob/master/etcs/experiments.md
add new mobilenet-v2 architecture.
Most helpful comment
https://github.com/ildoonet/tf-pose-estimation/blob/master/etcs/experiments.md
add new mobilenet-v2 architecture.