app.py

from sahi import utils, predict, AutoDetectionModel
from PIL import Image
import gradio as gr
import numpy
import torch

model_id_list = ['deprem-ml/Binafarktespit-yolo5x-v1-xview', 'SerdarHelli/deprem_satellite_labeled_yolov8', 'kadirnar/yolov7-v0.1', 'kadirnar/UNet-EfficientNet-b6-Istanbul']
current_device = "cuda" if torch.cuda.is_available() else "cpu"
model_types = ["YOLOv5", "YOLOv5 + SAHI", "YOLOv8", "YOLOv7"]

def sahi_yolov5_inference(
    image,
    model_id,
    model_type,
    image_size,
    slice_height=512,
    slice_width=512,
    overlap_height_ratio=0.1,
    overlap_width_ratio=0.1,
    postprocess_type="NMS",
    postprocess_match_metric="IOU",
    postprocess_match_threshold=0.25,
    postprocess_class_agnostic=False,
):
    
    rect_th = None or max(round(sum(image.size) / 2 * 0.0001), 1)
    text_th = None or max(rect_th - 2, 1)
    
    if model_type == "YOLOv5":
        # standard inference
        model = AutoDetectionModel.from_pretrained(
            model_type="yolov5",
            model_path=model_id,
            device=current_device,
            confidence_threshold=0.5,
            image_size=image_size,
            )
        
        prediction_result_1 = predict.get_prediction(
            image=image, detection_model=model
        )

        visual_result_1 = utils.cv.visualize_object_predictions(
            image=numpy.array(image),
            object_prediction_list=prediction_result_1.object_prediction_list,
            rect_th=rect_th,
            text_th=text_th,
        )
        
        output = Image.fromarray(visual_result_1["image"])
        return output
    
    elif model_type == "YOLOv5 + SAHI":
        model = AutoDetectionModel.from_pretrained(
            model_type="yolov5",
            model_path=model_id,
            device=current_device,
            confidence_threshold=0.5,
            image_size=image_size,
        )
        
        prediction_result_2 = predict.get_sliced_prediction(
            image=image,
            detection_model=model,
            slice_height=int(slice_height),
            slice_width=int(slice_width),
            overlap_height_ratio=overlap_height_ratio,
            overlap_width_ratio=overlap_width_ratio,
            postprocess_type=postprocess_type,
            postprocess_match_metric=postprocess_match_metric,
            postprocess_match_threshold=postprocess_match_threshold,
            postprocess_class_agnostic=postprocess_class_agnostic,
        )
        
        visual_result_2 = utils.cv.visualize_object_predictions(
            image=numpy.array(image),
            object_prediction_list=prediction_result_2.object_prediction_list,
            rect_th=rect_th,
            text_th=text_th,
        )
        
        output = Image.fromarray(visual_result_2["image"])
        return output

    elif model_type == "YOLOv8":
        from ultralyticsplus import YOLO, render_result

        model = YOLO('SerdarHelli/deprem_satellite_labeled_yolov8')
        result = model.predict(image, imgsz=image_size)[0]
        render = render_result(model=model, image=image, result=result)
        return render
    
    elif model_type == "YOLOv7":
        import yolov7
        
        model = yolov7.load(model_id, device="cuda:0", hf_model=True, trace=False)
        results = model([image], size=image_size)
        return results.render()[0]

    """
    elif model_type == "Unet-Istanbul":
        from istanbul_unet import unet_prediction
        
        output = unet_prediction(input_path=image, model_path=model_id)
        return output
    """
    
inputs = [
    gr.Image(type="pil", label="Original Image"),
    gr.Dropdown(choices=model_id_list,label="Choose Model",value=model_id_list[0]),
    gr.Dropdown( choices=model_types, label="Choose Model Type", value=model_types[1]),
    gr.Slider(minimum=128, maximum=2048, value=640, step=32, label="Image Size"),
    gr.Slider(minimum=128, maximum=2048, value=512, step=32, label="Slice Height"),
    gr.Slider(minimum=128, maximum=2048, value=512, step=32, label="Slice Width"),
    gr.Slider(minimum=0.0, maximum=1.0, value=0.1, step=0.1, label="Overlap Height Ratio"),
    gr.Slider(minimum=0.0, maximum=1.0, value=0.1, step=0.1, label="Overlap Width Ratio"),
    gr.Dropdown(["NMS", "GREEDYNMM"], type="value", value="NMS", label="Postprocess Type"),
    gr.Dropdown(["IOU", "IOS"], type="value", value="IOU", label="Postprocess Type"),
    gr.Slider(minimum=0.0, maximum=1.0, value=0.5, step=0.1, label="Postprocess Match Threshold"),
    gr.Checkbox(value=True, label="Postprocess Class Agnostic"),
]

outputs = [gr.outputs.Image(type="pil", label="Output")]

title = "Building Detection from Satellite Images with SAHI + YOLOv5"
description = "SAHI + YOLOv5 demo for building detection from satellite images. Upload an image or click an example image to use."
article = "<p style='text-align: center'>SAHI is a lightweight vision library for performing large scale object detection/ instance segmentation.. <a href='https://github.com/obss/sahi'>SAHI Github</a> | <a href='https://medium.com/codable/sahi-a-vision-library-for-performing-sliced-inference-on-large-images-small-objects-c8b086af3b80'>SAHI Blog</a> | <a href='https://github.com/fcakyon/yolov5-pip'>YOLOv5 Github</a> </p>"
examples = [
    ["data/26.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
    ["data/27.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
    ["data/28.jpg", 'deprem-ml/Binafarktespit-yolo5x-v1-xview', "YOLOv5 + SAHI", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
    ["data/31.jpg", 'deprem-ml/SerdarHelli-yolov8-v1-xview', "YOLOv8", 640, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False]
    #["data/Istanbul.jpg", 'kadirnar/UNet-EfficientNet-b6-Istanbul', "Unet-Istanbul", 512, 512, 512, 0.1, 0.1, "NMS", "IOU", 0.25, False],
]


demo = gr.Interface(
    sahi_yolov5_inference,
    inputs,
    outputs,
    title=title,
    description=description,
    article=article,
    examples=examples,
    theme="huggingface",
    cache_examples=True,
)

demo.launch(debug=True, enable_queue=True)