app.py

import gradio as gr
import numpy as np
from PIL import Image, ImageDraw, ImageFilter, ImageOps
from transformers import OneFormerProcessor, OneFormerForUniversalSegmentation, pipeline
import torchvision.transforms
import torch


# pipeline = pipeline(task="image-segmentation", model="shi-labs/oneformer_cityscapes_swin_large", label_ids_to_fuse=[11])

person_processor = OneFormerProcessor.from_pretrained("shi-labs/oneformer_cityscapes_swin_large")
person_model = OneFormerForUniversalSegmentation.from_pretrained("shi-labs/oneformer_cityscapes_swin_large")
transform = torchvision.transforms.ToPILImage()

def detect_person(image: Image):
    semantic_inputs = person_processor(images=image, task_inputs=["semantic"], return_tensors="pt")
    semantic_outputs = person_model(**semantic_inputs)
    predicted_semantic_map = person_processor.post_process_semantic_segmentation(semantic_outputs, target_sizes=[image.size[::-1]])[0]
    mask = transform(predicted_semantic_map.to(torch.uint8))
    mask = Image.eval(mask, lambda x: 0 if x == 11 else 255)
    return mask  


# def detect_person_pipeline():
#     results = pipeline(image, label_ids_to_fuse=[11])
#     for detection in results:
#         if detection["label"] == "person":
#             return ImageOps.invert(detection["mask"])
#     return Image.new(mode="L", size=image.size, color=255)


def detect_dummy(image: Image):
    return Image.new(mode="L", size=image.size, color=255)


detectors = {
    "Person": detect_person,
    # "License Plate": detect_license_plate
}

def test_gradio(image):
    masks = [detect_person(image)]
    combined = np.minimum.reduce([np.array(m) for m in masks])
    mask = Image.fromarray(combined)
    # Apply blur through mask
    blurred = image.filter(ImageFilter.GaussianBlur(15))
    anonymized = Image.composite(image, blurred, mask)
    return anonymized


demo_live = gr.Interface(
    fn=test_gradio, 
    inputs=gr.Image(source="webcam", type="pil", shape=(640, 480)), 
    outputs=gr.Image(type="pil")
)

demo_upload = gr.Interface(
    fn=test_gradio, 
    inputs=gr.Image(type="pil", shape=(640, 480)), 
    outputs=gr.Image(type="pil")
)

demo = gr.TabbedInterface(
    interface_list=[demo_live, demo_upload], 
    tab_names=["Webcam", "Bild hochladen"], 
    title="Image Anonymizer"
)

# if __name__ == "__main__":
    # demo.launch()
    # 

print(__name__)
demo.launch(server_name="localhost", server_port=8080)