# CardsClassification

**Repository Path**: baymax104/CardsClassification

## Basic Information

- **Project Name**: CardsClassification
- **Description**: Practice Project for Cards Image Dataset-Classification Dataset
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-06-17
- **Last Updated**: 2024-06-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 基于CNN的扑克牌图像分类

## 数据集介绍

数据集源自Kaggle的公开数据集[Cards Image Dataset-Classification](https://www.kaggle.com/datasets/gpiosenka/cards-image-datasetclassification)

数据集放置在根目录下，目录形式如下

- ML/dataset
  - test
    - ace of clubs
    - ...
  - train
    - ace of clubs
    - ...
  - valid
    - ace of clubs
    - ...
  - cards.csv

其中标签有53类，训练集图像7624张，验证集图像265张，测试集图像265张，类别分布较均衡

图像大小均为224\*224，包含RGB三通道

对图像进行归一化和转置后得
- 训练集Tensor(7624, 3, 224, 224)
- 验证集Tensor(265, 3, 224, 224)
- 测试集Tensor(265, 3, 224, 224)


## 网络构建

使用CNN+残差连接

<img src="https://typora-images-1309988842.cos.ap-beijing.myqcloud.com/img/image-20230601114904010.png" alt="image-20230601114904010" style="zoom:50%;" />

## 训练结果

训练验证结果

<img src="https://typora-images-1309988842.cos.ap-beijing.myqcloud.com/img/image-20230601114951076.png" alt="image-20230601114951076" style="zoom: 50%;" />

测试结果

<img src="https://typora-images-1309988842.cos.ap-beijing.myqcloud.com/img/image-20230601115012703.png" alt="image-20230601115012703" style="zoom:50%;" />