🐶🐱 Cat vs Dog Classifier API - FastAPI Deployment

A Deep Learning image classification model for Cats and Dogs using PyTorch and FastAPI, achieving around 60% accuracy. Users can upload images and receive classification results via a REST API.

---

🎯 Objectives

• Train a Cat/Dog image classification model using ResNet18 (pre-trained).

• Build a RESTful API with FastAPI to serve the model.

• Allow users to send images and get prediction results.

---

📁 Project Structure

catdog-classifier-inference-fastapi/
│
├── app.py                        # Initialize FastAPI app
├── server.py                     # Launch server with Uvicorn
├── requirements.txt              # Dependency list
├── animal_classification.ipynb   # Model training notebook
│
├── config/
│   ├── catdog_cfg.py             # Model & image preprocessing config
│   └── logging_cfg.py            # Logging config
│
├── logs/                         # Log files
│
├── middleware/
│   ├── __init__.py
│   ├── cors.py                   # CORS configuration
│   └── http.py                   # Log HTTP request
│
├── models/
│   ├── weights/
│   │   └── catdog_weights.pt     # Trained model weights
│   ├── catdog_model.py           # Model architecture
│   └── catdog_predictor.py       # Predictor class
│
├── routes/
│   ├── base.py                   # Route aggregator
│   └── catdog_route.py           # Route aggregator
│
├── schemas/
│   └── catdog_schema.py          # Response schema definition
│
└── utils/
    └── logger.py                 # Logger setup

---

🚀 Running Locally

1. Clone the Repository

git clone https://github.com/KhoiBui16/catdog-classifier-inference-fastapi.git
cd catdog-classifier-inference-fastapi

2. Create a Virtual Environment (Optional but Recommended)

Using venv:

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

Using conda:

conda create -n catdog_env python=3.10
conda activate catdog_env

3. Install Dependencies

pip install -r requirements.txt

4. Run the FastAPI Application

python server.py

• API will be hosted at: http://0.0.0.0:8000

• Swagger UI: http://localhost:8000/docs

---

🧠 Model Overview

• Architecture: ResNet18 pre-trained on ImageNet.

• Dataset: cats_vs_dogs from HuggingFace Datasets.

• Input: RGB image, resized to 64x64, normalized using ImageNet stats.

• Output: Softmax probabilities for Cat or Dog class.

• Accuracy: ~60% on validation set.

---

📸 API Endpoint

Method	Endpoint	Description
POST	/catdog_classification/predict	Upload an image and get prediction

Request:

• Type: multipart/form-data

• Field: file_upload (JPG, PNG...)

Response (JSON):

{
  "probs": [0.41, 0.59],
  "best_prob": 0.59,
  "predicted_id": 1,
  "predicted_class": "Dog",
  "predictor_name": "resnet18"
}

---

🧪 Testing the API

Once server.py is running:

1. Visit: http://localhost:8000/docs

2. Use the /catdog_classification/predict endpoint

3. Upload an image and click "Execute" to view the result

---

📚 References

• FastAPI Documentation
• PyTorch
• HuggingFace Datasets