luckycallor/InsightFace-tensorflow

Issues

• output node of the second pre-model(from baidu Drive, ａs if it were the first pre-trained model

  #39 opened 5 years ago by tjulitianyi1997
  1

• embedding comparision

  #48 opened 4 years ago by marziehoghbaie
  0

• 训练时Acc会一直为0吗？

  #46 opened 5 years ago by zch-ssc
  4

• 请问一下不同数据集训练一些超参数怎么设置的？

  #47 opened 4 years ago by Nakupenda-7
  0

• 训练模型的准确率

  #12 opened 6 years ago by destiny311
  3

• 请问backbone里的resnet_v1和v2以及modifiedv2有对比吗，哪个效果最好？

  #45 opened 5 years ago by FelixZhang7
  0

• 关于训练时loss的问题

  #44 opened 5 years ago by vvblack
  1

• FileNotFoundError:

  #33 opened 5 years ago by snow-wind-001
  1

• Ask about IDE

  #34 opened 5 years ago by ThanhPham1987
  1

• pre_trained models

  #11 opened 6 years ago by destiny311
  1

• questions about training dataset

  #42 opened 5 years ago by vvblack
  1

• 下载lfw官网数据集制成tfrecord后运行finetune_softmax.py时acc起始很低

  #43 opened 5 years ago by chenkeyi2013
  3

• finetune loss always big

  #14 opened 6 years ago by x3393467
  2

• mode of mxrec convert to tfrecord

  #28 opened 5 years ago by Deepcong2019
  3

• 1006k的预训练模型和334k的好像是同样一个压缩包

  #26 opened 5 years ago by sunruina2
  2

• questions about threshold

  #41 opened 5 years ago by vvblack
  1

• good job

  #40 opened 5 years ago by damaocun
  1

• checkpoint file so large(643M)

  #5 opened 6 years ago by clhne
  5

• When I want to use my own data to train the model, I first transfer the dataset to .tfrecord format. How can I set the validation data format in config file? Just use the .tfrecord or only use .bin?

  #37 opened 5 years ago by GodMon
  4

• 经测试效果不好，本项目还需要进一步完善！

  #38 opened 5 years ago by CachCheng
  0

• 多GPU训练，获取embedding的时候报错KeyError: "The name 'embeddings:0' refers to a Tensor which does not exist.

  #36 opened 5 years ago by HandsLing
  0

• Model for any other case is severely overfitted

  #35 opened 5 years ago by AkhilAshref
  0

• Question about: "face images should be well cropped here"

  #19 opened 5 years ago by khanhnt
  2

• Couple of errors in get_embd.py

  #32 opened 5 years ago by AkhilAshref
  0

• missing requirements

  #31 opened 5 years ago by SaltyHash123
  0

• 两张照片不同的人，距离也很小

  #29 opened 5 years ago by HandsLing
  2

• Extracting incorrect embedding in get_embd.py

  #30 opened 5 years ago by qwn19970728
  0

• re-implementation with tensorflow cannot achieve the same accuracy on mxnet.

  #27 opened 5 years ago by Jar7
  1

• 2 different faces but small distance?

  #15 opened 6 years ago by huytq97
  18

• should not only restore trainable variables in `get_embd.py`?

  #25 opened 5 years ago by HongtaoYang
  1

• Is there any problem about get_embd.py？

  #13 opened 6 years ago by YJingyu
  3

• Normalization of embeddings in evaluate.py

  #24 opened 5 years ago by AmitRozner
  0

• Questions about training

  #23 opened 5 years ago by christk1
  2

• 使用训练好的模型在faces_ms1m_112x112.tfrecord数据上验证时准确率很低？

  #22 opened 5 years ago by xuweihuawei
  4

• Why do you have configurations of different batch sizes?

  #18 opened 5 years ago by BruceDai003
  1

• Question: How do I generate a validation set .bin file with my own data?

  #20 opened 5 years ago by kingguan4
  1

• ValueError: dictionary update sequence element #0 has length 5; 2 is required

  #21 opened 5 years ago by 2151330
  1

• Why you dont add summary to summary writer?

  #17 opened 5 years ago by milk-bottle-liyu
  0

• How do I validate on my own validation set?

  #16 opened 6 years ago by Wenlin-Chen
  0

• how to compare to embeddins of faces for knowing that these faces are similar

  #8 opened 6 years ago by Muhammad333
  2

• Pre-trained models

  #1 opened 6 years ago by TimoBolkart
  3

• How exactly is the cropping done?

  #10 opened 6 years ago by lagidigu
  5

• distance_metric == 1,evaluate.py, when to choose 'distance_metric == 1'?

  #9 opened 6 years ago by Wxsang
  2

• valid dataset

  #2 opened 6 years ago by csqzs
  2

• questins about training time

  #6 opened 6 years ago by friedhelm739
  4

• About image normalization

  #4 opened 6 years ago by yinguobing
  2

• Low accuracy

  #3 opened 6 years ago by nnt296
  1