我如何在Tensorflow中准备自己的数据集（数千张图片）？

Question

I have some thousands of images in .PNG format. The labels for these png files are stored in txt file. PNG files names are like 1.png, 2.png, 3.png and so on... TXT file contains labels separated by new lines (Each line contains a single lable)

How can I prepare them for use in tensorflow?

Once prepared/loaded, I want to train a CNN in tensorflow by using the images mentioned. It will be highly appreciated If someone can give specific examples of reading png and txt files in tensorflow.

Answer 1

You might wanna check out tensorflow slim library, including its dataset class and use of data_set_data_provider.py

slim: https://github.com/tensorflow/tensorflow/tree/r0.11/tensorflow/contrib/slim

dataset: https://github.com/tensorflow/tensorflow/tree/r0.11/tensorflow/contrib/slim/python/slim/data