https://github.com/vickshan001/cifar-10-cnn-enhancer-neural-networks
CNN classifier for CIFAR-10 with enhanced architecture, dropout, and data augmentation.
https://github.com/vickshan001/cifar-10-cnn-enhancer-neural-networks
adam-optimizer cifar10 cnn deep-learning dropout image-classification neural-networks python xavier-initialization
Last synced: 6 months ago
JSON representation
CNN classifier for CIFAR-10 with enhanced architecture, dropout, and data augmentation.
- Host: GitHub
- URL: https://github.com/vickshan001/cifar-10-cnn-enhancer-neural-networks
- Owner: vickshan001
- Created: 2025-03-30T21:49:07.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-03-30T21:51:33.000Z (6 months ago)
- Last Synced: 2025-03-30T22:28:07.789Z (6 months ago)
- Topics: adam-optimizer, cifar10, cnn, deep-learning, dropout, image-classification, neural-networks, python, xavier-initialization
- Language: Jupyter Notebook
- Homepage:
- Size: 0 Bytes
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🧠 CIFAR-10 CNN Enhancer – Neural Networks
The aim was to improve the classification accuracy of a CNN model on the CIFAR-10 dataset through architecture tuning, data augmentation, and dropout regularization.
---
## 📦 Dataset
[CIFAR-10](https://www.cs.toronto.edu/~kriz/cifar.html) – a 60,000-image dataset across 10 classes like airplane, bird, cat, deer, dog, etc.
---
## 🧠 Key Enhancements
### 🔁 Data Augmentation
- `RandomHorizontalFlip()`
- `RandomCrop(32, padding=4)`
### 🧱 Model Architecture
- Intermediate block improvements:
- Dropout for regularization
- Adapted fully connected layers
- Output block:
- Multiple FC layers with ReLU activation
- Final FC layer outputs raw logits
### 🧪 Initialization & Optimisation
- Xavier (Glorot) initialization for weights
- **Adam Optimizer** with CrossEntropy Loss
---
## 📊 Training Results
- Accuracy increased gradually across epochs
- Final **Test Accuracy: 62%**
- Visualization of loss and accuracy over epochs
---
## 📂 Project Structure
- `Final_Score.ipynb` – Full notebook including architecture, training, and evaluation
---
## 🚀 How to Run
1. Clone the repository
2. Run `Final_Score.ipynb` in Jupyter Notebook
3. Required Libraries:
- `torch`, `torchvision`, `numpy`, `matplotlib`
---
## 🏫 Module Info
- 📅 Year: 2023/24
- 🏫 University: Queen Mary University of London
- 👨💻 Author: Vickshan Vicknakumaran
---
## 📜 License
For educational and research purposes only.