MASK-RCNN常见配置出错问题解决

tensorflow.python.framework.errors_impl.UnknownError: 2 root error(s) found.
  (0) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1/convolution}}]]
     [[rpn_bbox/concat/_4279]]
  (1) Unknown: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
     [[{{node conv1/convolution}}]]
0 successful operations.
0 derived errors ignored.

解决方法，添加如下代码即可：

import os

os.environ['CUDA_VISIBLE_DEVICES'] = '/gpu:0'

'Model' object has no attribute 'metrics_tensors' 问题解决！！！

self.keras_model.metrics_tensors = []

File "/home/liqiang/anaconda3/envs/tensorflow1.8/lib/python3.6/site-packages/keras/engine/saving.py", line 1004, in load_weights_from_hdf5_group
    original_keras_version = f.attrs['keras_version'].decode('utf8')
AttributeError: 'str' object has no attribute 'decode'

 卸载原来的h5py模块，安装2.10版本

pip install h5py==2.10 -i https://pypi.tuna.tsinghua.edu.cn/simple/

解决ImportError: Keras requires TensorFlow 2.2 or higher. Install TensorFlow via `pip install tensorflow

我的python是3.7版本，安装的tensorflow-gpu是1.15.0,对应的keras版本是2.3.1

(python37) C:\Users\admin>pip install keras==2.3.1 -i https://pypi.douban.com/simple