The project bases on keras to detect the driver's actions which are correct or dangerous, thank you for your reading The project will use the VGG , ResNet and MobileNet_v1 to detect the driver datasets. The project running the pycharm, so for the .sh file you should use the Git (recommend).
In the directory, you can get the VGG models ,ResNet models, and MobileNet_v1 based on the keras or tensorflow.
This is a result. when you run the code ,you can get the result in here.
There are your datasets
dowmloading from https://www.kaggle.com/competitions/state-farm-distracted-driver-detection/data.
This is a .sh file ,you should run it for training the models
This is a .sh file too, you can run it for testing the models.
The file is the important, you can change something you want.
In the here , you should pay attention to your datasets、models... 's file path!!!!
The file is about show many result after training the model and saving it.
This is about the process from data.
If you want to create a app or use MCU、CPU which is removable,you need run the code to transform your models.
The project is to predict a picture which on your computer and to show the result.
https://github.com/yeyupiaoling/ClassificationForAndroid/tree/master/TFLiteClassification
and
https://resource.doiduoyi.com/#c1uo2s4
The above is important for the app created, and we can run the two projects to load ourselves' models in the mobile phone.
我制作的app核心代码全部来自于这两个网址,处于诚实且保护原作者成果的原则,只贴附上原作者代码。
以下两点是其中修改符合自己模型的处理:
1. 第一个网址的代码中:注意自己tflite的输入层和输出层的名字是否一致数据部分。记得修改均值为0,方差为255f!(这里根据读者预处理决定)
2. 第二个网址的代码中:因为是10个类别,需要修改数组大小,本人网络训练预处理预处理用的是除以255f,因此这里也需要这么处理!
This is my first create project in the github, If you can see here ,thank you!!.
The project is my undergraduate graduation thesis's code, and I want to remenber the beautiful time.