这是最简单的CNN模型。
mnist_model.py
实现了 MNIST 的 Estimator 代码。
transfer_and_finetune.py
实现了基于前者训练好的参数,再进行迁移学习的 Estimator,代码中将其体现为迁移到Hnd首写字母训练集上
这是进阶的模型。
根目录有两个模型,分别是 Keras 实现的 Inception_v3 和 Tensorflow layers API 实现的 Inception_v3 的 Estimator。
keras_to_estimator
目录包括了将 Keras 的 Inception_v3 模型转换为 Estimator,迁移学习,保存和加载的方法。
这是再一步进阶的模型。
- 整理好代码,多写一些注释,上传
欢迎Star和Issue哇 ^3^