nyyznyyz1991

@KaleidoZhouYN I used FN to train, but the saved variance is NAN. In inference phase, the embedding feature( use_global_status = true) is used , after subtracting mean and dividing variance,...

@KaleidoZhouYN Thanks a lot, i have trained a 99.5% model using FN and sphereloss, there are some confusing questions: 1. using FN layer would save the trained mean and variance,...

Thanks for your contribution to TensorRT implementation @chenzhi1992 , I found that TensorRT3 already implemented createPReLU ` INvPlugin * createPReLUPlugin(float negSlope); INvPlugin * createPReLUPlugin(const void * data, size_t length);` However...

@KaleidoZhouYN Amazing! can you share the 28-layer mode prototxt and solver.prototxt? I have been struggling to train 64 layer resnet with large datasets. And what is the batch_size you set...

for 20 layers, my loss value is 1.x and lfw test accuracy is 99.22%

i followed the author's solver,prototxt, didn't change anything. 28000 iters base_lr: 0.1 display: 100 max_iter: 28000 lr_policy: "multistep" gamma: 0.1 momentum: 0.9 weight_decay: 0.0005 snapshot: 1000 snapshot_prefix: "result/sphereface_model" solver_mode: GPU...

@Astrosun Have you solved this pro yet? I also want to know :)

@Jerryzcn @catsdogone Could you share how to transfer model from tensorflow to caffe? rewrite caffe's prototxt from scratch based on tensorflow or write a transfer.py script to achieve it?