TensorFlow下使用SSD在windows中遇到的问题
问题一:训练速度很慢,没有使用GPU
解决:
安装匹配的GPU版本的TensorFlow。
问题二:InvalidArgumentError (see above for traceback): Default MaxPoolingOp only supports NHWC on device type CPU
解决:
修改train_ssd_network.py文件中27行。
DATA_FORMAT = 'NHWC' # NCHW error: Invalid argument: Default MaxPoolingOp only supports NHWC on device type CPU
参考:
InvalidArgumentError (see above for traceback): Default MaxPoolingOp only supports NHWC on device ty
I have the following error when training SSD-Tensorflow. How can I solve it?
问题三:InvalidArgumentError (see above for traceback): Assign requires shapes of both tensors to match. lhs shape= [12] rhs shape= [84]
解决:
预测时定义的网络结构的参数和训练时保存的网络结构参数不匹配,修改train_ssd_network.py文件,使测试时参数与训练模型时相同。
参考:
InvalidArgumentError (see above for traceback): Assign requires shapes of both tensors to match. lhs