CIFAR-10 Image Classifier API

A REST API that classifies images into 10 categories using a Convolutional Neural Network (CNN) trained on the CIFAR-10 dataset. Built with TensorFlow and served via FastAPI.

---

Demo Results

Real-world images tested against the live API:

Image	Prediction	Confidence	Result
red car.jpeg	automobile	99.54%	✅ Correct
airplane flying.jpeg	airplane	96.90%	✅ Correct
ship at sea.jpeg	truck	62.35%	❌ Misclassified

Why the ship was misclassified: CIFAR-10 resizes all images to 32×32 pixels before classification. At that resolution, a large container ship with a boxy hull closely resembles a truck. This is a known limitation of low-resolution datasets — the model lacks the spatial detail needed to distinguish between similarly shaped objects.

---

Tech Stack

Layer	Technology
Model	TensorFlow / Keras CNN
API	FastAPI
Server	Uvicorn
Dataset	CIFAR-10 (60,000 images, 10 classes)
Language	Python 3.10+

---

CIFAR-10 Classes

The model classifies images into one of these 10 categories:

airplane · automobile · bird · cat · deer · dog · frog · horse · ship · truck

---

Project Structure

cifar10-api/
├── main.py              # FastAPI app and /predict endpoint
├── model.py             # CNN model architecture and training script
├── cifar10_model.h5     # Trained model weights
├── requirements.txt     # Python dependencies
└── README.md

---

Getting Started

1. Clone the Repository

git clone https://github.com/ShivamKumar20-AI/cifar10-api.git
cd cifar10-api

2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Run the API

uvicorn main:app --reload

The API will be live at http://127.0.0.1:8000

---

API Usage

Endpoint

POST /predict

Request

Send a multipart/form-data request with your image file:

curl -X POST "http://127.0.0.1:8000/predict" \
     -H "accept: application/json" \
     -H "Content-Type: multipart/form-data" \
     -F "file=@/path/to/your/image.jpg"

Response

{
  "prediction": "automobile",
  "confidence": "99.54%",
  "all_scores": {
    "airplane": "0.00%",
    "automobile": "99.54%",
    "bird": "0.00%",
    "cat": "0.00%",
    "deer": "0.00%",
    "dog": "0.00%",
    "frog": "0.00%",
    "horse": "0.00%",
    "ship": "0.40%",
    "truck": "0.05%"
  }
}

Interactive Docs

FastAPI auto-generates interactive documentation. Once the server is running, visit:

• Swagger UI: http://127.0.0.1:8000/docs
• ReDoc: http://127.0.0.1:8000/redoc

---

Model Architecture

The CNN was trained from scratch on the CIFAR-10 training set:

• Input: 32×32 RGB images
• Architecture: Convolutional layers → MaxPooling → Dropout → Dense layers → Softmax output
• Training: 10 epochs on 50,000 training images
• Validation: Tested on 10,000 held-out images

---

Limitations

• Images are resized to 32×32 pixels before classification — fine detail is lost
• The model was trained on the standard CIFAR-10 dataset; domain-specific imagery (e.g., satellite photos, medical images) will not perform well
• Classes outside the 10 CIFAR-10 categories will be forced into the nearest match

---

Author

Shivam Kumar — AI & ML Engineer