https://github.com/abdul-rafay19/cnnect-classifier
A real-time image classification app built with PyTorch and deployed using Streamlit. CNNect-Classifier uses a custom-trained CNN on the CIFAR-10 dataset to predict image classes with confidence scores. Designed to demonstrate end-to-end deep learning deployment — from training to user-facing web app.
https://github.com/abdul-rafay19/cnnect-classifier
ai artificial-intelligence artificial-intelligence-algorithms artificial-neural-networks cifar-10 cifar10 cnn cnn-classification cnn-model cnnect deep-learning machine-learning neural-network neural-networks numpy python pytorch pytorch-cnn streamlit streamlit-webapp
Last synced: 6 months ago
JSON representation
A real-time image classification app built with PyTorch and deployed using Streamlit. CNNect-Classifier uses a custom-trained CNN on the CIFAR-10 dataset to predict image classes with confidence scores. Designed to demonstrate end-to-end deep learning deployment — from training to user-facing web app.
- Host: GitHub
- URL: https://github.com/abdul-rafay19/cnnect-classifier
- Owner: abdul-rafay19
- Created: 2025-04-12T08:46:13.000Z (6 months ago)
- Default Branch: main
- Last Pushed: 2025-04-12T08:57:53.000Z (6 months ago)
- Last Synced: 2025-04-13T11:14:28.257Z (6 months ago)
- Topics: ai, artificial-intelligence, artificial-intelligence-algorithms, artificial-neural-networks, cifar-10, cifar10, cnn, cnn-classification, cnn-model, cnnect, deep-learning, machine-learning, neural-network, neural-networks, numpy, python, pytorch, pytorch-cnn, streamlit, streamlit-webapp
- Language: Jupyter Notebook
- Homepage:
- Size: 1.91 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# CNNect-Classifier
**Abdul Rafay** | BS Software Engineering
CNNect-Classifier is a deep learning-powered web application that performs real-time image classification on the CIFAR-10 dataset. Built as part of my Machine Learning Internship at Intern Intelligence, this project represents a complete end-to-end pipeline — from model training to deployment — wrapped inside a user-friendly interface.
This project started as a task but turned into something much bigger. Inspired by a post that said Streamlit is the fastest way to build AI apps and impress future employers, I took that challenge seriously. I moved beyond notebooks and built a fully functional deep learning app.
## What It Does
- Takes an input image via upload
- Runs it through a custom-trained CNN model
- Outputs predicted class with probability scores
- Provides a clean, mobile-friendly UI for interaction
## Behind the Scenes
- Built a CNN from scratch using PyTorch for image classification
- Used convolutional layers, pooling, dropout, and fully connected layers
- Trained and fine-tuned the model on CIFAR-10 for optimal accuracy
- Switched from TensorFlow to PyTorch for smoother experimentation
- Deployed using Streamlit to create a real-time web interface
## Tech Stack
- PyTorch
- Torchvision
- Streamlit
- PIL
- NumPy
- HTML/CSS
## Features
- 📷 Image Upload Interface
- ⚡ Real-Time Predictions
- 📊 Class-wise Confidence Breakdown
- 💻 Mobile-Responsive, Minimalist UI
## Installation
To run the app locally, follow these steps:
```bash
git clone https://github.com/YourGitHubUsername/CNNect-Classifier.git
cd CNNect-Classifier
streamlit run app.py
```
Make sure you have the trained model file `cifar10_model.pt` in the same directory as `app.py`.
## File Structure
- `app.py` – Streamlit frontend with prediction logic
- `task.ipynb` – Model building and training notebook
- `cifar10_model.pt` – Saved PyTorch model weights (not included in repo)
## Classes Predicted
- Airplane
- Automobile
- Bird
- Cat
- Deer
- Dog
- Frog
- Horse
- Ship
- Truck
## Why This Project Matters
This wasn’t just about training a model — it was about making machine learning usable. I learned how to bridge the gap between model performance and user experience. CNNect-Classifier represents my journey into real-world AI application development, with hands-on deployment and user-facing design.
## Connect with Me
📌 [LinkedIn – Abdul Rafay](https://www.linkedin.com/in/abdul-rafay19)