Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/fanhuaandluomu/imageclassification
原文:http://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html
https://github.com/fanhuaandluomu/imageclassification
image-classification keras
Last synced: 16 days ago
JSON representation
原文:http://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html
- Host: GitHub
- URL: https://github.com/fanhuaandluomu/imageclassification
- Owner: FanhuaandLuomu
- Created: 2017-08-03T07:05:16.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2017-08-04T00:27:56.000Z (over 7 years ago)
- Last Synced: 2024-11-12T11:12:05.913Z (3 months ago)
- Topics: image-classification, keras
- Language: Python
- Size: 8.79 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# ImageClassification
原文:http://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html
参考上述面向小数据集的图像分类方法(kaggle的猫狗分类),重新实现了可运行的py程序,对一些细节做了改动。
# base model
从图片中直接训练一个小网络(作为基准方法)
python base_cnn_model.py
va_acc:0.81
# vgg16提取特征+训练全连接分类
利用预训练网络vgg16的bottleneck(瓶颈)特征,输入至全连接网络,sigmoid分类
python vgg16_feature.py
val_acc:0.90
# fine_tune vgg16
fine-tune预训练网络(vgg+全连接sigmoid)的高层,冻结网络的前几层(卷积特征提取层)
python vgg16_fine_tune.py
val_acc:0.95+# 语料下载
为方便操作,语料我已整理出实验所需格式,
其中train:1000+1000, validation:400+400
百度网盘下载:http://pan.baidu.com/s/1qY6tc4G