🚗 Car Brand Classification using Deep Learning & FastAPI

A complete machine learning pipeline for classifying car brand images using PyTorch, deployed via FastAPI, and tested under load using Locust.

🎥 Demo Video: https://youtu.be/QTSoxvG2vEM

🌐 API URL (Local or Cloud) http://127.0.0.1:8000/predict

Local URL: http://localhost:8501 Network URL: http://192.168.1.78:8501

📌 Project Overview

This project implements an end-to-end image classification solution:

✔ Dataset Preprocessing

Folder-based dataset (brand → images)

Augmentation: resize, normalize, flip, rotation

Train/Test loaders

✔ Model Training

PyTorch CNN or ResNet

Training loop with accuracy & loss tracking

Trained model saved as best_model.pth

✔ FastAPI Deployment

REST endpoint for image prediction

Upload image → returns predicted car brand

Fully tested using Postman and Swagger UI

✔ Load Testing with Locust

Simulates flood requests

Measures RPS, latency, failure rate

Ensures API scalability

✔ Notebook Included

Contains full workflow from preprocessing → training → testing.

🛠️ Installation & Setup 1️⃣ Clone the Repository git clone (https://github.com/JaboJean/Car_Classification_ml.git) cd car_classification_ml

2️⃣ Install Dependencies pip install -r requirements.txt

Or using Conda:

conda create -n carml python=3.10 conda activate carml pip install -r requirements.txt

3️⃣ Start the FastAPI Server uvicorn src.api:app --reload

Open Swagger UI:

http://127.0.0.1:8000/docs

4️⃣ Make a Prediction

Use Swagger UI or send an image via Python:

from predict import predict_image print(predict_image("path/to/car.jpg"))

🧪 Flood Request / Load Testing (Locust) Run Locust locust -f locustfile.py

Dashboard:

http://localhost:8089

📘 Jupyter Notebook Contents

My notebook includes:

📌 1. Data Preprocessing

Image transforms

Data visualization

Train/test split

📌 2. Model Training

CNN / ResNet architecture

Loss & accuracy tracking

Saved model weights

📌 3. Testing & Prediction

Evaluation metrics

Confusion matrix

Single-image prediction function

📌 4. Model File

Stored here:

saved_models/car_model.pth

📂 Project Structure car_classification_ml/ │── src/ │ ├── api.py # FastAPI app │ ├── model.py # CNN/ResNet model │ ├── predict.py # Prediction logic │── notebook/ │ ├── car_classification.ipynb │── saved_models/ │ ├── best_model.pth │── data/ │ ├── train/ │ ├── test/ │── locustfile.py │── requirements.txt │── README.md