Skin Disease Identification using CNNs
muneeb arshad
Skin Disease Identification using Convolutional Neural Networks
Overview
This project was completed as part of my Final Year Project (FYP).
The objective was to classify skin diseases from dermatoscopic images using deep learning.
We explored different dataset preparation techniques and compared performance across two datasets and two models.
Datasets
Self-Collected Dataset
Images collected manually.
Dataset was augmented using techniques such as:
Augmentation was crucial due to the limited number of original images.
Google Images Dataset
Additional dataset curated from Google Images.
Used to increase variety and size of training samples.
Models
Custom CNN Model
Designed and implemented from scratch.
Consists of convolutional, pooling, and dense layers.
Keras Built-in Model
Transfer learning approach using a pre-trained CNN.
Fine-tuned on our datasets.
Study & Experimentation
We conducted experiments using two datasets and two models, resulting in four combinations:
Custom CNN + Self-Collected Dataset
Custom CNN + Google Images Dataset
Keras Pretrained Model + Self-Collected Dataset
Keras Pretrained Model + Google Images Dataset
Each combination was evaluated based on accuracy, loss curves, and generalization ability.
Results
Data Augmentation improved the accuracy of the self-collected dataset.
Transfer Learning (Keras pretrained model) consistently outperformed the custom CNN.
Project Report
The complete FYP Report with detailed methodology, experiments, and analysis is available in a separate repository: https://github.com/muneebsaddal/Final-Year-Project-report/
Like this project
Posted Oct 16, 2025
Classified skin diseases using CNNs with data augmentation and transfer learning.
Likes
0
Views
0
