ailias/Focal-Loss-implement-on-Tensorflow

Issues

• Focal loss with MobileNet-SSD FPNLite.

  #16 opened 2 years ago by Annieliaquat
  1

• ∵0**0=1 ∴ gamma != 0

  #15 opened 4 years ago by zhangyuqi-1
  2

• 因为0**0=1，所以需要增加γ=0时的特殊情况。否则会出现问题

  #14 opened 4 years ago by zhangyuqi-1
  0

• 多分类时参数α的意义是什么？

  #13 opened 5 years ago by ChuanTianML
  0

• Focal loss value is very less

  #2 opened 7 years ago by prakashjayy
  6

• The Code can be try in multi label ?

  #8 opened 6 years ago by hecongqing
  2

• train my data

  #12 opened 5 years ago by henbucuoshanghai
  0

• tarin my data???????how???????

  #11 opened 5 years ago by henbucuoshanghai
  0

• Question on MobileNet SSD architechture with focal loss.

  #5 opened 6 years ago by MAGI003769
  4

• Why add "neg_p_sub" into "per_entry_cross_ent"

  #10 opened 6 years ago by CoderChang
  1

• Should the tf.nn.sigmoid be replaced with tf.nn.sotfmax

  #9 opened 6 years ago by mzhaoshuai
  7

• How to implemented it in faster-rcnn?

  #6 opened 6 years ago by zqdeepbluesky
  4

• what does array_ops.where mean?

  #7 opened 6 years ago by DRosemei
  1

• tf.nn.sigmoid and tf.nn.softmax diff?

  #4 opened 7 years ago by ronyuzhang
  1

• The issue about @slim.add_arg_scope

  #3 opened 7 years ago by zzzzzz0407
  1

• hi,are you sure your focal loss equals to -alpha*(1-pt)^gamma*log(pt)

  #1 opened 7 years ago by li10141110
  1