yizt/keras-faster-rcnn

Issues

• loss出现Nan改了很久，希望您帮忙看一下

  #45 opened 4 years ago by Wangcancan1127
  0

• 训练出错

  #44 opened 4 years ago by wudi00
  0

• 训练成功，想咨询您预测的问题

  #43 opened 4 years ago by Wangcancan1127
  0

• 训练过程中,训练loss一开始是有效的,然后变成nan,只训练了1000左右次nan后又变为正常值,很神奇，感觉内部有个bug。

  #13 opened 5 years ago by 836304831
  12

• 自己的数据集。 在数据加载的时候。出现这个问题‘NoneType' object has no attribute 'find'’。 验证集和测试集加载不了

  #42 opened 4 years ago by MARUIHAN123
  0

• 你好，loss一直为nan，自己的数据集。什么问题呀

  #41 opened 4 years ago by MARUIHAN123
  1

• TypeError: Incompatible types: <dtype: 'resource'> vs. int64. Value is 0

  #40 opened 4 years ago by MARUIHAN123
  3

• TypeError: Incompatible types: <dtype: 'resource'> vs. int32. Value is 0

  #29 opened 5 years ago by shadow12138
  2

• @MARUIHAN123 就是1.19.0

  #39 opened 4 years ago by MARUIHAN123
  1

• @dengchubo 我用的是TensorFlow 1.14.0

  #38 opened 4 years ago by MARUIHAN123
  1

• rcnn_class_loss训练到一半就会变成nan

  #37 opened 4 years ago by huangxudong
  7

• 您好，如果想预测出测试集图片后把所有的预测结果图片保存到一个文件夹，不是随机展示，该怎么改呢

  #36 opened 4 years ago by Wangcancan1127
  2

• 您好，预训练模型连接失效了，可以更新一下吗，感谢

  #35 opened 4 years ago by Wangcancan1127
  9

• 执行train.py一直OOM

  #32 opened 5 years ago by BngThea
  11

• 没有画出框

  #34 opened 4 years ago by huangxudong
  1

• mAP過低

  #33 opened 5 years ago by enoke
  0

• 请问如何训练不带gtbox的图片

  #31 opened 5 years ago by ashipiling
  4

• 坐标归一化

  #30 opened 5 years ago by Crispus1119
  5

• 为啥我聚类聚出负数来了

  #28 opened 5 years ago by shadow12138
  1

• acc

  #27 opened 5 years ago by liby542698
  1

• 报错 'Model' object has no attribute 'add_metric'

  #26 opened 5 years ago by dengchubo
  2

• 请问我该怎样把basenet改成VGG16呢

  #25 opened 5 years ago by Faterjames
  2

• 训练的时候感觉loss也很低，大概在2左右，但是做evaluate mAP 基本是0 。

  #18 opened 5 years ago by sibadakesi
  5

• epoch

  #23 opened 5 years ago by liby542698
  2

• 有人对这个代码增加FPN吗？

  #24 opened 5 years ago by sibadakesi
  1

• 训练完了没有发现已经保存的 model

  #22 opened 5 years ago by hzcforever
  0

• 运行 inference.py 后程序终止

  #21 opened 5 years ago by hzcforever
  0

• rcnn_miss_gt_num非常高？ 有没有什么解决办法

  #20 opened 5 years ago by sibadakesi
  3

• 关于map_fn

  #19 opened 5 years ago by sibadakesi
  1

• RPN网络的class输出没有经过softmax直接进行计算交叉熵损失了？

  #17 opened 5 years ago by dishofchicken
  1

• 请问这个可以训练自己的数据集吗，该修改哪些地方吗

  #2 opened 6 years ago by huangzhenjie
  7

• 我的rcnn_class_loss是NAN，是不是每个image裡面一定都要至少要有一个object？

  #15 opened 5 years ago by derekhsu
  4

• evaluate error

  #16 opened 5 years ago by NiFangBaAGe
  3

• 运行inference.py进程异常退出

  #14 opened 5 years ago by 836304831
  1

• use tf.keras instead of keras run the code encounting error

  #12 opened 5 years ago by 836304831
  4

• 在train.py中 m = models.frcnn(config, stage='train') 这一行报错

  #8 opened 6 years ago by bendaxi
  6

• voc_path = '/opt/dataset/VOCdevkit'目錄找不到？

  #11 opened 6 years ago by kaonick
  2

• 训练报错

  #10 opened 6 years ago by hjgherman
  5

• NAN

  #7 opened 6 years ago by wudi00
  9

• 点赞支持

  #9 opened 6 years ago by qingmeizhujiu
  1

• How about the performance？

  #1 opened 6 years ago by bityangke
  6

• 运行错误

  #6 opened 6 years ago by wudi00
  2

• train.rcnn.log

  #5 opened 6 years ago by wudi00
  2

• 训练计算的mAP

  #4 opened 6 years ago by anshan3
  2

• tensorflow.python.framework.errors_impl.NotFoundError: Resource __per_step_8/_tensor_arraysrpn2proposals/map/TensorArray_2_17/N10tensorflow11TensorArrayE does not exist.

  #3 opened 6 years ago by yizt
  0