A flexible and extensible framework for gait recognition. You can focus on designing your own models and comparing with state-of-the-arts easily with the help of OpenGait.
664
stars
154
forks
source link
Why batch shuffle affects deepgaitv2 training results #217
Batch_shuffle simply disrupts the order of person IDs, and when calculating triple loss, all embeddings are put together, Shuffle has no effect on the calculation of triple loss. Will softmax be affected? If it doesn't affect why there will be differences in training results. Did I misunderstand somewhere.Thank you for your help.
Batch_shuffle simply disrupts the order of person IDs, and when calculating triple loss, all embeddings are put together, Shuffle has no effect on the calculation of triple loss. Will softmax be affected? If it doesn't affect why there will be differences in training results. Did I misunderstand somewhere.Thank you for your help.