luhaofang/tripletloss

Issues

• 您好，我想用自己的数据集跑这个网络，除了修改config路径，还需要修改哪里？

  #61 opened 7 years ago by gomeansai
  2

• 训练到后面 loss都变为0.1 an,ap变为0

  #46 opened 8 years ago by Johere
  6

• shuffle

  #52 opened 6 years ago by ucasqcz
  0

• ap,an都变得特别大

  #64 opened 6 years ago by MarStarck
  0

• 想问一下大家的batch_size都设多大？

  #32 opened 8 years ago by DaleAG
  4

• 工作推荐

  #63 opened 6 years ago by rongbo
  0

• tipletselectlayer - computing the distance against the anchor image

  #51 opened 7 years ago by identidata
  2

• Does tripletselectlayer have bug?

  #27 opened 9 years ago by wangxianliang
  1

• Online triplet generation

  #62 opened 7 years ago by amena6490
  1

• 请问FaceNet使用的训练集从哪里下载

  #35 opened 8 years ago by soldier828
  1

• 数据集组织需要什么特别处理么?

  #59 opened 7 years ago by Sunwzh
  1

• online triplet sample selection usage?

  #60 opened 7 years ago by Phoebe-star
  0

• what does no_residual_list for?

  #58 opened 7 years ago by Lowtec-Sam
  0

• Negative mining in TripletSelectLayer

  #57 opened 7 years ago by PramuPerera
  1

• fc9_1 weights is increase with the train iterations from 0.0x to 40.x

  #53 opened 7 years ago by zhouyongxiu
  1

• Use of the margin

  #56 opened 7 years ago by gombru
  2

• how to train it on 2 gpus

  #55 opened 7 years ago by marcosly
  0

• Training your code on custom dataset

  #54 opened 7 years ago by abhisheksgumadi
  1

• 你好，我用你的代码训练卡住了，卡在learning rate policy :step那里，不知道怎么进行下去，求指导，方便的话加下我qq，798638754

  #34 opened 8 years ago by poorfriend
  2

• No module named norm2layer

  #22 opened 9 years ago by zhangfudiyi
  15

• the triplet loss architecture is unsupervised, is this correct?

  #50 opened 7 years ago by identidata
  0

• Hard Sample

  #49 opened 7 years ago by ccJia
  0

• training problem

  #48 opened 7 years ago by ccJia
  0

• training problem

  #47 opened 7 years ago by peternara
  5

• loss 在0与0.0x之前变化

  #30 opened 8 years ago by twinsyssy1018
  9

• layer_params = yaml.load(self.param_str_) AttributeError: 'DataLayer' object has no attribute 'param_str_'

  #28 opened 9 years ago by lcj1105
  10

• Selecting semi-hard examples, possible bug?

  #37 opened 8 years ago by guyov1
  0

• the module name is right,but it still failed with boost::python::error_already_set

  #45 opened 8 years ago by lemonaha
  0

• tripletloss 改善了几个点？

  #25 opened 9 years ago by dianxin556
  33

• 为什么在embedding后面接norm层，不是在norm层后面接embedding

  #43 opened 8 years ago by gaojie0105
  2

• AttributeError: 'module' object has no attribute 'text_format'

  #44 opened 8 years ago by Jamie0813
  1

• deploy.prototxt question

  #42 opened 8 years ago by ZoroDerVonCodier
  2

• gradient formulation for unit normalized Image embedding

  #41 opened 8 years ago by nitneh
  0

• Question about Anchors

  #40 opened 8 years ago by ZoroDerVonCodier
  2

• tripletselectlayer possible bug?

  #39 opened 8 years ago by TuBui
  1

• 遇到一个很奇怪的错误，请大家看一下

  #33 opened 8 years ago by louielu1027
  1

• pos, neg从大到小排序，这里有bug吗

  #36 opened 8 years ago by zuowang
  1

• 您好，您这个工程到底怎么用起来啊？有详细的步骤吗？

  #31 opened 8 years ago by louielu1027
  3

• 没有hard positive和hard negative会不会很慢？

  #29 opened 8 years ago by lliuz
  0

• Data Layer Creating Problem

  #24 opened 9 years ago by zhaoxin19
  18

• train_va.txt的格式

  #23 opened 9 years ago by wuhuaming2006
  1

• Hi，我想问下您的997个人的数据 每个人样本有多少 样本数差距大么 另外样本里面侧脸多吗？

  #16 opened 9 years ago by yzk0281
  1

• loss居高不下

  #18 opened 9 years ago by zhangfudiyi
  8

• tripletselectlayer.py第58行是不是有问题？

  #26 opened 9 years ago by chendd14
  1

• Hi tripletloss finetune的时候 怎样判断已经finetune好了？

  #15 opened 9 years ago by yzk0281
  17

• 请教一个浅显的问题

  #19 opened 9 years ago by thu-zxs
  2

• 叨扰了，再请教一个问题

  #20 opened 9 years ago by thu-zxs
  0

• training image size ??

  #21 opened 9 years ago by peternara
  0

• Hi, What have you extract feature?

  #17 opened 9 years ago by peternara
  4

• Hi 请问你放出来的模型是自己从头开始训练的还是在官方的模型基础上finetune的？

  #14 opened 9 years ago by yzk0281
  1