WitchPuff / CRC-Diagnose

A medical image recognition project powered by self-implemented ResNet and ViT models, utilizing PyTorch, with a user-friendly web demo built on Flask and Vue. 该项目使用PyTorch开发,基于自主实现的ResNet和ViT模型及NCT-CRC-HE-100K数据集,实现了直结肠癌医学图像识别,配备了基于Flask和Vue的用户友好型Web演示。

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Questions about performance results

LightersWang opened this issue · comments

Hi!

Thanks for your great work in this repo.

I'm curious about the accuracy of your provided ResNet and ViT on the NCT-CRC-HE-100K dataset after training with your codebase.

Hi :D
Sure, here's the performance after 40 epochs:
ResNet: 1 for trainset, 0.93 for testset
ViT: 0.929 for trainset, 0.886 for testset

Hope it helps!

Thanks for your help!