search

yangxue0827 / FPN_Tensorflow

A Tensorflow implementation of FPN detection framework.
415 stars 150 forks source link

convert_data_to_tfrecord.py error! #22

Open lfdeep opened 6 years ago

lfdeep commented 6 years ago

Hello,i run convert_data_to_tfrecord.py in python 3.6! Traceback (most recent call last): File "convert_data_to_tfrecord.py", line 159, in convert_pascal_to_tfrecord() File "convert_data_to_tfrecord.py", line 140, in convert_pascal_to_tfrecord 'img_name': _bytes_feature(img_name), File "convert_data_to_tfrecord.py", line 28, in _bytes_feature return tf.train.Feature(bytes_list=tf.train.BytesList(value=[value])) TypeError: '11222.jpg' has type str, but expected one of: bytes

yangxue0827 commented 6 years ago

'img_name': _bytes_feature(img_name.encode())

lfdeep commented 6 years ago

Thank you!You are right, I solved problems in your way. And following: 2018-06-09 06:39:43.143172: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA restore model it is running in CPU,how to change to GPU?

lfdeep commented 6 years ago

@yangxue0827 您好，上面问题已解决。请问一下eval.py代码中怎么给真实标签呢？我现在有训练好的模型了，想直接得到AP和mAP

yangxue0827 commented 6 years ago

@lfdeep 你好，不建议用这个代码里的评估函数，有问题。你可以用faster作者写的评估函数，我也会在以后根新评估代码，现在太忙没有时间。

lfdeep commented 6 years ago

@yangxue0827 谢谢，那test.py和train.py有问题吗？能生成结果的txt吗?