import supervision as sv import gradio as gr from ultralytics import YOLO import sahi import numpy as np # Images sahi.utils.file.download_from_url( "https://www.erbanotizie.com/wp-content/uploads/2014/01/Casello.jpg", "ocr1.jpg", ) sahi.utils.file.download_from_url( "https://media-cdn.tripadvisor.com/media/photo-s/15/1d/03/18/receipt.jpg", "ocr2.jpg", ) sahi.utils.file.download_from_url( "https://upload.forumfree.net/i/ff11450850/b5ef33b7-01da-4055-9ece-089b2a35a193.jpg", "ocr3.jpg", ) annotatorbbox = sv.BoxAnnotator() annotatormask=sv.MaskAnnotator() model = YOLO("best_Receipt.pt") def yolov8_inference( image: gr.inputs.Image = None, conf_threshold: gr.inputs.Slider = 0.5, iou_threshold: gr.inputs.Slider = 0.45, ): image=image[:, :, ::-1].astype(np.uint8) model = YOLO("https://huggingface.co/spaces/devisionx/first-demo/blob/main/best_Receipt.pt") results = model(image,imgsz=320,conf=conf_threshold,iou=iou_threshold)[0] image=image[:, :, ::-1].astype(np.uint8) detections = sv.Detections.from_yolov8(results) annotated_image = annotatormask.annotate(scene=image, detections=detections) annotated_image = annotatorbbox.annotate(scene=annotated_image , detections=detections) return annotated_image ''' image_input = gr.inputs.Image() # Adjust the shape according to your requirements inputs = [ gr.inputs.Image(label="Input Image"), gr.Slider( minimum=0.0, maximum=1.0, value=0.25, step=0.05, label="Confidence Threshold" ), gr.Slider(minimum=0.0, maximum=1.0, value=0.45, step=0.05, label="IOU Threshold"), ] outputs = gr.Image(type="filepath", label="Output Image") title = "OCR Demo" ''' examples = [ ["ocr1.jpg", 0.6, 0.45], ["ocr2.jpg", 0.25, 0.45], ["ocr3.jpg", 0.25, 0.45], ] outputs_images = [ ["1.jpg"], # First example: an output image for the cat example ["2.jpg"] # Second example: an output image for the dog example ,["3.jpg"] ] readme_html = """

More details:

We present a demo for performing object segmentation using a model trained on OCR-Receipt dataset. The model was trained on 54 training images and validated on 15 images.

Usage:

You can upload receipt images, and the demo will provide you with your segmented image.

Dataset:

This dataset comprises a total of 77 images, which are divided into three distinct sets for various purposes:

License: This dataset is made available under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

To access and download this dataset, please follow this link: Dataset Download

""" with gr.Blocks() as demo: gr.Markdown( """

OCR Demo

Powered by Tuba
""" ) # Define the input components and add them to the layout with gr.Row(): image_input = gr.inputs.Image() outputs = gr.Image(type="filepath", label="Output Image") # Define the output component and add it to the layout with gr.Row(): conf_slider=gr.Slider(minimum=0.0, maximum=1.0, value=0.25, step=0.05, label="Confidence Threshold" ) with gr.Row(): IOU_Slider=gr.Slider(minimum=0.0, maximum=1.0, value=0.45, step=0.05, label="IOU Threshold") button = gr.Button("Run") # Define the event listener that connects the input and output components and triggers the function button.click(fn=yolov8_inference, inputs=[image_input, conf_slider,IOU_Slider], outputs=outputs, api_name="yolov8_inference") gr.Examples( fn=yolov8_inference, examples=examples, inputs=[image_input, conf_slider,IOU_Slider], outputs=[outputs] ) # gr.Examples(inputs=examples, outputs=outputs_images) # Add the description below the layout gr.Markdown(readme_html) # Launch the app demo.launch(share=False)