Skip to content

Semantic Segmentation

Model basic information

A lightweight portrait segmentation model based on the ExtremeC3 model. For more details, please refer to: ExtremeC3_Portrait_Segmentation project.

Reference：https://github.com/PaddlePaddle/PaddleHub/blob/release/v2.2/modules/image/semantic_segmentation/ExtremeC3_Portrait_Segmentation

Sample result example

Enter the file path and the model will give its predictions：

image

image

image

Let's try it out now

Prerequisite

1、environment dependent

Please visit dependencies

2、ExtremeC3_Portrait_Segmentation dependent

  • paddlepaddle >= 2.0.0

  • paddlehub >= 2.0.0

3、Download the model

hub install ExtremeC3_Portrait_Segmentation

Serve the Model

Install Pinferencia

First, let's install Pinferencia.

pip install "pinferencia[streamlit]"

Create app.py

Let's save our predict function into a file app.py and add some lines to register it.

app.py
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
import base64
from io import BytesIO

import cv2
import numpy as np
import paddlehub as hub
from PIL import Image

from pinferencia import Server, task

semantic_segmentation = hub.Module(name="ExtremeC3_Portrait_Segmentation")


def base64_str_to_cv2(base64_str: str) -> np.ndarray:
    return cv2.imdecode(
        np.fromstring(base64.b64decode(base64_str), np.uint8), cv2.IMREAD_COLOR
    )


def predict(base64_img_str: str) -> str:
    images = [base64_str_to_cv2(base64_img_str)]
    result = semantic_segmentation.Segmentation(
        images=images,
        output_dir="./",
        visualization=True,
    )
    pil_img = Image.fromarray(result[0]["result"])
    buff = BytesIO()
    pil_img.save(buff, format="JPEG")
    return base64.b64encode(buff.getvalue()).decode("utf-8")


service = Server()
service.register(
    model_name="semantic_segmentation",
    model=predict,
    metadata={"task": task.IMAGE_TO_IMAGE},
)

Run the service, and wait for it to load the model and start the server:

$ uvicorn app:service --reload
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)
INFO:     Started reloader process [xxxxx] using statreload
INFO:     Started server process [xxxxx]
INFO:     Waiting for application startup.
INFO:     Application startup complete.

$ pinfer app:service --reload

Pinferencia: Frontend component streamlit is starting...
Pinferencia: Backend component uvicorn is starting...

Test the service

Open http://127.0.0.1:8501, and the template Url Image To Image will be selected automatically.

png

Request

curl --location --request POST \
    'http://127.0.0.1:8000/v1/models/semantic_segmentation/predict' \
    --header 'Content-Type: application/json' \
    --data-raw '{
        "data": "/9j/4AAQSkZJRgABAQEA/..."
    }'

Response

{
    "model_name": "semantic_segmentation",
    "model_version": "default",
    "data": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRo..."
}

Create the test.py.

test.py
1
2
3
4
5
6
7
8
9
import requests


response = requests.post(
    url="http://localhost:8000/v1/models/semantic_segmentation/predict",
    headers = {"Content-type": "application/json"},
    json={"data": "/9j/4AAQSkZJRgABAQEA/..."},
)
print(response.json())
Run the script and check the result.

$ python test.py
{
    "model_name": "semantic_segmentation",
    "model_version": "default",
    "data": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRo..."
}

Even cooler, go to http://127.0.0.1:8000, and you will have a full documentation of your APIs.

You can also send predict requests just there!