Skip to content

An image classification neural network using multi scale features.

Notifications You must be signed in to change notification settings

jeongahyun/CombNet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 

Repository files navigation

CombNet: Diagnosis of Skin Diseases using Multi-Scale Features

result

This is a Pytorch implementation of CombNet. (paper here)
CombNet is a neural network created for image classification: conceived in the feature pyramid structure. I used datasets from kaggle - Skin Cancer MNIST: HAM10000


Architecture

architecture

Backbone of CombNet is made up of ResNet-18, supported by two types of sub-network from different scale feature maps.
The following formula is used to calculate the total loss from the three losses:

equation

In this experiment, the best result was when alpha = 0.5, betha = 0.7
From this model, you can get around 5% higher accuracy than the plain ResNet accuracy!

acc

Environment

The model is trained using following hardware:

  • GTX TITAN X (Pascal) - 12GB VRAM
  • Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz
  • 64GB RAM

The code is developed under the following software:

  • Ubuntu 16.04.6 LTS
  • CUDA V10.1.243
  • Python 3.6.10
  • PyTorch 1.5.0

About

An image classification neural network using multi scale features.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages