本文主要通过改进后的YOLOv5模型结合Mask R-CNN模型实现外周血细胞的计数任务,但是由于两部分算法模型是分开的两个python项目操作起来较繁琐,所以编写一个桌面应用将二者整合起来。 开发环境为Python 3.6(conda),编译器Pycharm CE 2021,程序开发工具为tkinter。
(1)输入:在程序主窗口输入所要处理的血细胞图片路径
![image](https://private-user-images.githubusercontent.com/115158062/332028965-96fae405-0058-491b-9985-9f27f9c7496e.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAxMTEyNzcsIm5iZiI6MTcyMDExMDk3NywicGF0aCI6Ii8xMTUxNTgwNjIvMzMyMDI4OTY1LTk2ZmFlNDA1LTAwNTgtNDkxYi05OTg1LTlmMjdmOWM3NDk2ZS5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwNzA0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDcwNFQxNjM2MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT01OTg4ODZhYTViNjAxODA5MDY5NDNjZTc0ZGY0YjVmMzY5M2UxZjUzZTYyMWNkOGM0OTgzNjg1MjJmY2UyMWI5JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.y2gTohEb3UEBBzbaGQ88FOwwmxwKirtCSKNW15sY1Dw)
(2)YOLOv5目标检测:将已经训练好的权重文件和预测代码嵌入程序中,点击按钮【YOLO】完成目标检测操作,并生成带有Bounding boxes的图片,不同类别血细胞数量信息将会显示在窗口右侧。
![image](https://private-user-images.githubusercontent.com/115158062/332028995-6e5f7b35-e3b4-44b0-8aec-d39a12ad2397.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAxMTEyNzcsIm5iZiI6MTcyMDExMDk3NywicGF0aCI6Ii8xMTUxNTgwNjIvMzMyMDI4OTk1LTZlNWY3YjM1LWUzYjQtNDRiMC04YWVjLWQzOWExMmFkMjM5Ny5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwNzA0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDcwNFQxNjM2MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT02YTM3NmRkZmI4NWFkYjNiMzQzNmJiYWRlYTAyNzI4NDg1MzBjZjNiN2I2MmNjOGM2Y2VmNWYzOTYzMzgzY2EyJlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.l_WS_JLAUl-xsZToMCxu-uCwx9bE2MjaX_WBMQqoCQs)
![image](https://private-user-images.githubusercontent.com/115158062/332029030-a9cf6b45-b7c9-4948-a368-64aa76463a30.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAxMTEyNzcsIm5iZiI6MTcyMDExMDk3NywicGF0aCI6Ii8xMTUxNTgwNjIvMzMyMDI5MDMwLWE5Y2Y2YjQ1LWI3YzktNDk0OC1hMzY4LTY0YWE3NjQ2M2EzMC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwNzA0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDcwNFQxNjM2MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT0yYWExZDJhYjQzZThlYWQxMmVkMzg0ZWNhMDBjNDg4Y2Y5YWFkZjNmZTk3ZTlkZjMzNDg4NGYxN2JlODM2Yjg4JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.c2NwnwnMrX4MjnM00J87P2eNMqVNlUPh_WhhUGO2ZdA)
(3)粘连细胞实例分割:将训练好的Mask R-CNN权重以及相关文件嵌入,点击【Mask】按钮,对之前检测到的每一个RBC2区域进行实例分割,并且可以查看每一个区域的实例分割带掩码图片,比如下图4个按钮对应4个重叠的红细胞,点击【0】即可查看第1个重叠红细胞实例分割后的图片。
![image](https://private-user-images.githubusercontent.com/115158062/332029263-25898723-ac30-4bc1-988c-9bf86b6e8fb8.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAxMTEyNzcsIm5iZiI6MTcyMDExMDk3NywicGF0aCI6Ii8xMTUxNTgwNjIvMzMyMDI5MjYzLTI1ODk4NzIzLWFjMzAtNGJjMS05ODhjLTliZjg2YjZlOGZiOC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwNzA0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDcwNFQxNjM2MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT1jOTA1NGY4YjBjNWNlNDE2ZjU4NGVlZWY0MzZkNDJiYjRhOWMwMDAyNmVmYTY3MmI0NTNjMjJjYTkwMjhkOTA4JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.8hgNJ-dHg6D-9fipDaBSd1J5IJdkcudrfBqzKm-Ta04)
![截屏2024-05-25 13 34 43](https://private-user-images.githubusercontent.com/115158062/333774896-6e863a8d-0b12-4b28-9dcc-beef88010acd.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MjAxMTEyNzcsIm5iZiI6MTcyMDExMDk3NywicGF0aCI6Ii8xMTUxNTgwNjIvMzMzNzc0ODk2LTZlODYzYThkLTBiMTItNGIyOC05ZGNjLWJlZWY4ODAxMGFjZC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwNzA0JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDcwNFQxNjM2MTdaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT03NTNiMmU1ZGJhNGU5YjhmNjQ4NWE4MDZkMjAzZTcxYjdhN2Y0OTcxMTA1N2I1ZmI3YjM5ZDBjYzg5ZjNiY2U2JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.9n20j-965Tilq8ofzSAqST6jrWA419mk8x31WBcMzvA)