Final validation and training loss?

mks0601 / TF-SimpleHumanPose

TensorFlow implementation of "Simple Baselines for Human Pose Estimation and Tracking", ECCV 2018

335 stars 90 forks source link

Final validation and training loss? #64

Closed wmcnally closed 3 years ago

wmcnally commented 4 years ago

Do you have on hand the final validation/training losses? I'm still struggling to reproduce the 70.4 AP with TF2. Highest I've gotten so far is 68.7... My validation loss is around 30. Thanks.

mks0601 commented 4 years ago

Sorry I don't have it :( I think the network architecture is quite simple, so data pre-processing such as affine transformation can matter.

wmcnally commented 4 years ago

Ok. No problem! Thanks for the quick reply.