InceptionV3_TensorFlow

InceptionV3_TensorFlow is an implementation of inception v3 using tensorflow and slim according to our guidline.

Dependencies

• TensorFlow (>= 0.12)

Features

• train
• predict
• save checkpoint
• real time data augumentation

Quick start

If you want a quick start to run training of Inception_v3, you can simply do:

./train.sh

The above script has passed test under Ubuntu15.10, CentOS and macOS.

If you want to go through the train process step by step, please take the following content as example.

Setup

1. download data in data/readme.md
2. execute "data/create_examples_list.py"
3. execute "data/relation_tag_to_id.py"
4. you can see train_csv.txt and test_csv.txt

Start to train##

python trainer.py

Pass test under Ubuntu15.10 and CentOS

How to use your own data sets

• create train_csv.txt and test_csv.txt in data directory.

datalist format

<image path>,<label number>  
...

• change num_classes in settings.py

Fine tune

• change fine_tune in settings.py

TensorBoard

tensorboard --logdir /to/your/path/train_dir --port=6006

Copyright (c) 2016 Masahiro Imai, Yixuan Hu (yeephycho) Released under the MIT license