About the creator

Sameed Ahmed

Student at LJMU • Creator of SkinSight

SkinSight is an AI-powered skin condition identifier built to support quick triage, improve accessibility to dermatology insights, and help users understand “what to do next”.

I have made this project Public on Github. Use the button below to download it. If you use or refer this project somewhere cite it as (Sameed LJMU Final Project, 2026).

GitHub LinkedIn BSCS Final Project
Sameed Ahmed

Technical specification

Key project choices across data, ML, and engineering.

Django + PyTorch + timm
Dataset
Kaggle Dermnet Dataset

by Shubham Goel & Bill Hall.

Training images
Trained On

15,500+ images

Base model
EfficientNetV2-S

Fast baseline using timm pretrained weights.

Design
Tailwind (CDN)

Modern UI with strong readability skin themed

Frontend
HTML, CSS, JS

Django templating language

Backend
Django

User accounts + private history + CSV export + model inference.

Database
PostgreSQL

SQLite3 was used during development and was later migrated to PostgreSQL for Deployment.

Deployment
Render

Django Skinsight Web-App was Deployed using render.

Domain & DNS Management
Namecheap

The domain skinsight.online was registered at Namecheap.

Model Training Process

How the SkinSight AI model was trained and fine-tuned for 23 skin condition categories.

Transfer learning + GPU training

1️⃣ Dataset Preparation

  • 🔹 Images organized by class using ImageFolder format.
  • 🔹 23 disease categories stored in class_names.json.
  • 🔹 Data split into training, validation, and test sets.
  • 🔹 Images resized to match model input resolution.

2️⃣ Preprocessing & Augmentation

  • 🔹 Resize to 320×320 to capture fine skin texture.
  • 🔹 Training augmentations: flips, color adjustments, MixUp & CutMix.
  • 🔹 ImageNet normalization for pretrained model compatibility.
  • 🔹 Validation/test use only resize + normalization.

3️⃣ Model Architecture

  • 🔹 EfficientNetV2-S from the timm library.
  • 🔹 Initialized with pretrained ImageNet weights.
  • 🔹 Final classification layer adapted for 23 classes.
  • 🔹 Trained using GPU acceleration for deeper fine-tuning.

4️⃣ Fine-tuning & Optimization

  • 🔹 Two-stage training: freeze backbone → full fine-tuning.
  • 🔹 Loss: SoftTargetCrossEntropy (supports MixUp training).
  • 🔹 Optimizer: AdamW with cosine learning-rate scheduling.
  • 🔹 Best model saved to artifacts/skinsight_model.pt.

Confusion matrix

Model evaluation

A table showing how many times each true class was predicted as each class (correct vs misclassified counts).

Confusion matrix
Normalized Confusion Matrix

The same table but converted to percentages, showing the proportion of predictions for each class instead of raw counts.

Confusion matrix

Justification

Why these technical choices were made.

Advanced architecture

EfficientNetV2-S provides strong accuracy while remaining computationally efficient, making it well suited for multi-class dermatological image classification.

High-resolution learning

Larger input images help capture fine skin texture and lesion details, which are critical for distinguishing visually similar conditions.

Transfer learning + fine-tuning

Starting from pretrained ImageNet weights and progressively fine-tuning the network improves performance on limited medical data and stabilizes training.

GPU-accelerated training

GPU training enables higher-resolution images and deeper optimization, significantly improving model learning compared to CPU-only training.

User-focused results

Clinical-style outputs such as urgency level, guidance, and red-flag indicators make predictions easier to interpret for non-experts.

For More Information

Check the final project report submitted on Canvis.

Future vision

Where SkinSight could go next.

Roadmap
  1. 1
    Better data + balancing

    More diverse images, class balancing, and cleaner labels.

  2. 2
    Stronger evaluation + training

    Top-k accuracy, calibration, and per-class performance dashboards. Training on powerful GPUs to improve accuracy.

3
Mobile-first UX
UX

Guided capture + a better camera flow.

4
Progress tracking
Habit

Timeline view + reminders to re-scan.

5
Clinician mode
Pro

Export reports + structured notes.

6
Final goal
Responsible AI

A responsible, explainable assistant that helps users decide whether to self-care or seek professional medical advice.