Semantic Segmentation API

Overview

This API leverages DeepLabV3+ with a ResNet50 backbone to perform binary semantic segmentation on plant-specific input images. The segmentation model implementation can be found here. Developed using FastAPI, the API accepts base64-encoded plant images, processes them through the model, and returns the resulting binary semantic segmentation mask encoded as base64.

Getting Started

Prerequisites

Python 3.9 or higher
PIL
FastAPI
Uvicorn
Gunicorn
PyTorch 2.2.1

Installation

Clone the repository

git clone https://github.com/mukund-ks/Backend-SSPSDLFAM

Change directory

cd Backend-SSPSDLFAM

Create a 'weights' directory under 'src' and place your model weights there

mkdir -p src/weights
cp /path/to/your_weights.pth src/weights/

Install dependencies

pip install -r requirements.txt

Usage

Runnning the API

To start the FastAPI server, run the following command:

uvicorn main:app --host 127.0.0.1 --port 8000

The API will be accessible at http://localhost:8000. Open your web browser or use tools like cURL or Postman to interact with API endpoints.

Running through Docker

Build the Docker Image:
```
docker build -t backend-sspsdlfam .
```
Run the Docker Container:
```
docker run -p 8000:8080 backend-sspsdlfam
```
The API will be accessible at http://localhost:8000.

Segmentation Endpoint

Method: POST
Path: /segment
Parameters:

{
    "img_base64": "base64-encoded image string"
}

Response

Upon successfull processing, the API responds with a JSON object containing the base64-encoded binary mask.

{
    "mask":"base64-encoded mask string"
}

Testing with Python

To test the API using Python, you can use the provided example script. The script loads an image, encodes it into base64, sends a POST request to the API's segmentation endpoint, and decodes the resulting base64-encoded segmentation mask.

import requests
import base64
from io import BytesIO
from PIL import Image

img_size = (256,256) # H,W of input image

with open("test_img.png", "rb") as image_file:
    encoded_img = base64.b64encode(image_file.read()).decode("utf-8")

response = requests.post(url="http://localhost:8000/segment", json={"img_base64": encoded_img})

json_res = response.json()

base64_mask = json_res["mask"]
mask_bytes = base64.b64decode(base64_mask)

mask_img = Image.open(BytesIO(mask_bytes))

mask_img.save("mask.png", format="PNG")

Authentication

No authentication is currently required for this API.

Contributing

Contributions are welcomed! Feel free to submit issues, feature requests, or pull requests.

Note

This project is part of the Capstone Project (Team-34) for college.

Chirs0503/Backend-SSPSDLFAM