seathiefwang / rankpose Goto Github PK

I The loss function of nn.MarginRankingLoss is "loss(x1,x2,y)=max(0,−y∗(x1−x2)+margin) " where y expect 1 or -1, but labels only has value in 0 or 1

when i tried to train the model in this repo, loss value is nan. Is there something wrong in my config file?

here is my training config :

# coding:utf-8
Net:
  net_type: 'ResNet'
  n_class: 3
Data:
  train_dir: "../data"
  val_dir: "../data"
  train_name: '300w_lp_for_rank.txt'
  val_name: 'aflw2000_filename.txt'
  train_type: 'RANK_300W'
  val_type: 'AFLW2000'
  target_size: 224
Train:
  max_epoch: 80
  batch_size: 64
  num_workers: 6
  test_every: 1
  resume: False
  pretrained_path:
  use_bined: False
  use_rank: True
Loss:
  loss_type: 'RANK'
Optimizer:
  mode: 'adam'
  base_lr: 0.0005
  t_max: 10

thanks a lot in advance.

Validation is done in 1969 images of AFLW2000 (not 2000 in paper).

Hello, in your paper you use l1 ranking loss, but your code is l2 ranking loss, can you explain that.
BTW, I have tried both and l2 outperforms l1 little.

Where is test.py?