app.py

import os
os.system('python -m pip install \'git+https://github.com/facebookresearch/detectron2.git\'')

import cv2
import gradio as gr

from detectron2.config import get_cfg
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog


def predict(input_image):
    # Initialise model
    cfg = get_cfg()
    cfg.merge_from_file(model_zoo.get_config_file("COCO-PanopticSegmentation/panoptic_fpn_R_101_3x.yaml"))
    cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-PanopticSegmentation/panoptic_fpn_R_101_3x.yaml")
    cfg.MODEL.DEVICE = "cpu"
    predictor = DefaultPredictor(cfg)

    assert input_image.shape[2] == 3
    height, width, _ = input_image.shape

    # Apply Panoptic segmentation
    panoptic_seg, segments_info = predictor(input_image)["panoptic_seg"]
    v = Visualizer(input_image[:, :, ::-1], MetadataCatalog.get(cfg.DATASETS.TRAIN[0]), scale=1.2)
    out = v.draw_panoptic_seg_predictions(panoptic_seg.to("cpu"), segments_info)
    segmented_image = out.get_image()[:, :, ::-1]

    # Resize image if required
    if not segmented_image.shape[:2] == (height, width):
        segmented_image = cv2.resize(segmented_image, (height, width))

    # Combine the segmented and original image
    combined_image = cv2.hconcat([segmented_image, input_image])

    return combined_image


# Create Gradio interface
image_input = gr.Image(type="pil", label="Input Image")

iface = gr.Interface(fn=predict,
                     inputs=[image_input],
                     outputs=gr.Image(type="pil"),
                     examples=[["examples/trump-fake.jpeg"]])
iface.launch()