app.py

import gradio as gr
import cv2
import tempfile
from ultralytics import YOLOv10
import pandas as pd
from gradio import processing_utils

df = pd.read_csv('image_class.csv')
df = df[['name', 'class']]
df.drop_duplicates(inplace=True)
print(len(df))
df1 = pd.read_csv('image_class.csv')
df1 = df1[['name', 'class', 'im_file']]
df1['file_name'] = df1['im_file'].apply(lambda v: v.split('_')[-1].split('.')[0])
df1.drop(columns='im_file', inplace=True)
df1.drop_duplicates(inplace=True)
print(df1)
print(len(df1))

def yolov10_inference(image, video, image_size, conf_threshold, iou_threshold):
    model = YOLOv10('./drug_yolov10.pt')
    # model = YOLOv10('./pills_yolov10.pt')
    if image:
        results = model.predict(source=image, imgsz=image_size, conf=conf_threshold, iou=iou_threshold)
        annotated_image = results[0].plot()
        # Print the detected objects' information (class, coordinates, and probability)
        box = results[0].boxes
        cls = [int(c) for c in box.cls.tolist()]
        cnf = [round(f,2) for f in box.conf.tolist()]
        clcf = '\n'.join([f'Class:{cls[i]} , Confidence:{cnf[i]*100}%' for i in range(len(cls))]) #list(zip(cls,cnf))
        name = '\n'.join([df[df['class']==n]['name'].item() for n in cls])
        file_name = image.split('_')[-1].split('.')[0]
        print(f'file name: {file_name}')
        try:
            drug_name  = df1[df1['file_name']==file_name]['name'].item()
            drug_class = df1[df1['file_name']==file_name]['class'].item()
            drug_name = f'{drug_class}, {drug_name}'
            print(drug_name)
        except:
            drug_name = 'No have data'
        # print(cls)
        # print(name)
        # print(type(clcf))
        # print("Object type:", box.cls)
        # print("Coordinates:", box.xyxy)
        # print("Probability:", box.conf)
        # print('box.class data tyupe', type(box.cls.tolist()))
        return annotated_image[:, :, ::-1], None, clcf, name, file_name, drug_name
    else:
        video_path = tempfile.mktemp(suffix=".webm")
        with open(video_path, "wb") as f:
            with open(video, "rb") as g:
                f.write(g.read())

        cap = cv2.VideoCapture(video_path)
        fps = cap.get(cv2.CAP_PROP_FPS)
        frame_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
        frame_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

        output_video_path = tempfile.mktemp(suffix=".webm")
        out = cv2.VideoWriter(output_video_path, cv2.VideoWriter_fourcc(*'vp80'), fps, (frame_width, frame_height))

        while cap.isOpened():
            ret, frame = cap.read()
            if not ret:
                break

            results = model.predict(source=frame, imgsz=image_size, conf=conf_threshold, iou=iou_threshold)
            annotated_frame = results[0].plot()
            out.write(annotated_frame)

        cap.release()
        out.release()

        return None, output_video_path


def yolov10_inference_for_examples(image, image_size, conf_threshold, iou_threshold):
    annotated_image, _, output_class, output_name = yolov10_inference(image, None, image_size, conf_threshold, iou_threshold)
    return annotated_image#, None, output_class, output_name

def app():
    with gr.Blocks():
        with gr.Row():
            with gr.Column():
                # image = gr.Image(type="pil", label="Image", visible=True)
                image = gr.Image(type="filepath", label="Image", visible=True)
                video = gr.Video(label="Video", visible=False)
                input_type = gr.Radio(
                    choices=["Image", "Video"],
                    value="Image",
                    label="Input Type",
                )
                file_name = gr.Textbox(label='File Name')
                file_name.change(outputs=file_name)
                drug_name = gr.Textbox(label='Drug Name (Ground Truth)')
                drug_name.change(outputs=drug_name)

                image_size = gr.Slider(
                    label="Image Size",
                    minimum=0,
                    maximum=1280,
                    step=10,
                    value=640,
                )
                conf_threshold = gr.Slider(
                    label="Confidence Threshold",
                    minimum=0.0,
                    maximum=1.0,
                    step=0.05,
                    value=0.25,
                )
                iou_threshold = gr.Slider(
                    label="IOU Threshold",
                    minimum=0,
                    maximum=1,
                    step=0.1,
                    value=0.6,
                )
                yolov10_infer = gr.Button(value="Detect Objects")
            
            with gr.Column():
                output_image = gr.Image(type="numpy", label="Annotated Image", visible=True)
                output_video = gr.Video(label="Annotated Video", visible=False)
                output_name = gr.Textbox(label='Predicted Drug Name')
                output_name.change(outputs=output_name)
                output_class = gr.Textbox(label='Predicted Class')
                output_class.change(outputs=output_class)

        def update_visibility(input_type):
            image = gr.update(visible=True) if input_type == "Image" else gr.update(visible=False)
            video = gr.update(visible=False) if input_type == "Image" else gr.update(visible=True)
            output_image = gr.update(visible=True) if input_type == "Image" else gr.update(visible=False)
            output_video = gr.update(visible=False) if input_type == "Image" else gr.update(visible=True)
            print(f'updated image: {image}')

            return image, video, output_image, output_video

        input_type.change(
            fn=update_visibility,
            inputs=[input_type],
            outputs=[image, video, output_image, output_video],
        )

        def run_inference(image, video, image_size, conf_threshold, iou_threshold, input_type):
            if input_type == "Image":
                return yolov10_inference(image, None, image_size, conf_threshold, iou_threshold)
            else:
                return yolov10_inference(None, video, image_size, conf_threshold, iou_threshold)


        yolov10_infer.click(
            fn=run_inference,
            inputs=[image, video, image_size, conf_threshold, iou_threshold, input_type],
            outputs=[output_image, output_video, output_class, output_name, file_name, drug_name],
        )

        gr.Examples(
            examples = [
            ['./RXBASE-600_00071-1014-68_NLMIMAGE10_5715ABFD.jpg', 280, 0.2, 0.6],
            ['./RXNAV-600_13668-0095-90_RXNAVIMAGE10_D145E8EF.jpg', 640, 0.2, 0.7],
            ['./RXBASE-600_00074-7126-13_NLMIMAGE10_C003606B.jpg', 640, 0.2, 0.8],
            ],
            fn=yolov10_inference_for_examples,
            inputs=[
                image,
                image_size,
                conf_threshold,
                iou_threshold,
            ],
            outputs=[output_image],
            cache_examples='lazy',
        )

gradio_app = gr.Blocks()
with gradio_app:
    gr.HTML(
        """
    <h1 style='text-align: center'>
    YOLOv10: Real-Time End-to-End Object Detection
    </h1>
    """)
    gr.HTML(
        """
        <h3 style='text-align: center'>
        <a href='https://arxiv.org/abs/2405.14458' target='_blank'>arXiv</a> | <a href='https://github.com/THU-MIG/yolov10' target='_blank'>github</a>
        </h3>
        """)
    with gr.Row():
        with gr.Column():
            app()
if __name__ == '__main__':
    gradio_app.launch()