# python-utils

**Repository Path**: yuchende/python-utils

## Basic Information

- **Project Name**: python-utils
- **Description**: 本仓库用来存储，作者自己打包的工具包
- **Primary Language**: Python
- **License**: MulanPSL-2.0
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 2
- **Forks**: 1
- **Created**: 2020-06-15
- **Last Updated**: 2021-04-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# python-utils

#### Introduction
this repository use to calculate the metric of Semantic Segmentation


#### Installation

        pip install git+https://gitee.com/yuchende/python-utils.git@master

#### Example
        
        import metric
        import metric.calMetric as mc
        import torch
        
        # init
        train_acc = 0
        train_miou = 0
        train_class_acc = 0
        train_pixel_accuracy = 0

        calmc = mc.CalMetricUtils(5) # 语义分割中对应着5个分类，mask分别是 0,1,2,3,4

        # you net such as unet ,fcn8s .......
        out = net(train_image)                     # out  为 tensorSize([2,5,4,4])
        prelabel = out.max(1)[1].data.cpu().numpy()# prelabel shape = 2 * 1 * 4 * 4
        prelabel = [i for i in prelabel]           # prelabel shape = 2 * 4 * 4
        
        truelabel = imglabel.data.cpu().numpy()    # imglabel 为 tensorSize([2,1,4,4])
        truelabel = [i for i in truelabel]         # truelabel shape = 2 * 4 * 4
        
        # begin to calculate
        resultMetric = calmc(pre_label,true_label) # 
        
        train_acc  += resultMetric['mean_class_accuracy'] 
        train_miou += resultMetric['miou']
        train_class_acc += resultMetric['class_accuracy']
        train_pixel_accuracy += resultMetric['pixel_accuracy']
        
        # 说明
        # 当前计算的是一个批次的准确率，如果有n个批次，迭代之后，准确率计算如下
        # eg:
        #    train_acc / n