Image classification practise

Build the custom_train_dataset

The first step is to:build image classification dataset, partition training set and test set,collect image,download example data set,delete more unuseful file,do image size statics,proportional distribution,take_photos distribution,all kinds of image data.

Install config environment

one is to build config on local environment.and the other is to use cloud environment. In my personal view,it depends on what you focus on more.For me,I focus on how to use pytorch not how to install it.
The most convinent way is to use GPU cloud platform.

Collect image

1. Use tech of spyder to collect image.
   Don’t be fear how to use it.Just imitate other’s guy code. that’s ok!
   Remember take prefessional guy’s thing for your temperature use.

2. mark image classification dataset
   First:image dataset should constains as many as possible situatiions.
3. delete useless image directory and file.

Demo

1.use wget to download dataset directly.
2.do statics image size and distribution.

Do trainning set and test set classification.

This is important and difficulty point.
Make the image in directory visiable.
Do statics of various kinds of image classification dataset.

Models for train image recognize ways

1.No Code :Platform: paddle ModelArts
2.Code: package:pytorch tensorflow