app.py

#app7.py
import gradio as gr
import torch
from PIL import Image
from torchvision import transforms
from ultralyticsplus import YOLO, render_result
# import matplotlib.pyplot as plt
import numpy as np

torch.hub.download_url_to_file(
    'https://i.postimg.cc/g2xGJ4Qs/NSTA-Test-IMG-3276.jpg', 'NSTA.jpg')
torch.hub.download_url_to_file(
    'https://i.postimg.cc/BZCSwj2T/NSTB-Test-IMG-1472.jpg', 'NSTB.jpg')
torch.hub.download_url_to_file(
    'https://i.postimg.cc/yYY1q7Tw/NSTC-Test-IMG-0118.jpg', 'NSTC.jpg')
torch.hub.download_url_to_file(
    'https://i.postimg.cc/zD9ZQX6z/KCCA-Test-IMG-3555.jpg', 'KCCA.jpg')
torch.hub.download_url_to_file(
    'https://i.postimg.cc/vZLPXP7L/KCCB-Test-IMG-3733.jpg', 'KCCB.jpg')
torch.hub.download_url_to_file(
    'https://i.postimg.cc/BZFYqFmF/KCCC-Test-IMG-3892.jpg', 'KCCC.jpg')

def detect_objects(image_path, selected_model):
    # Open the image file and resize it
    image = Image.open(image_path)
    resized_image = image.resize((1024, 768))

    #default
    # model_path = ('MvitHYF/v8mvitcocoaseed2024')
    
    # Load the model
    nstcurrentmodel = "NST Model"
    kcccurrentmodel = "KCC Model"
    currentmodel = str(selected_model)
    nstcurrentmodel = "NST Model"
    if currentmodel == nstcurrentmodel:
        print("this is nst model")
        #best of NST Model
        #model_path = ('code/runs/train45/best.pt')
        model_path = ('MvitHYF/v8mvitcocoaseed2024')
    elif currentmodel == kcccurrentmodel:
        print("this is kcc model")
        #model_path = ('code/runs/kcc/v8/train83/best.pt')
        model_path = ('MvitHYF/kccv8mvitcocoaseed2024')

    #model_path = ('code/runs/train45/best.pt')
    # model_path = ('MvitHYF/v8mvitcocoaseed2024')
    model = YOLO(model_path)
    # model = YOLO('MvitHYF/v8mvitcocoaseed2024')

    # Set model parameters
    model.overrides['conf'] = 'null'  # NMS confidence threshold
    model.overrides['iou'] = 0.70  # NMS IoU threshold
    model.overrides['agnostic_nms'] = True  # NMS class-agnostic
    model.overrides['max_det'] = 1000  # maximum number of detections per image
    
    # Perform inference
    results = model.predict(resized_image)

    #debug check count
    # print("see")
    # print(results)
    cls = results[0].boxes.cls
    # print(cls)
    strcls = str(cls)
    # print(type(strcls))
    # print(strcls)
    count_classa = strcls.count('0')
    # print('Count of classA:', count_classa)
    count_classb = strcls.count('1')
    # print('Count of classB', count_classb)
    count_classc = strcls.count('2')
    # print('Count of classC:', count_classc)
    intcount_classa = int(count_classa)
    intcount_classb = int(count_classb)
    intcount_classc = int(count_classc)
    total = intcount_classa + intcount_classb + intcount_classc 
    # print("end see")

    # gr.Image(label="Pie Graph")
    # Format the output to print the counts
    output_counts = f"Totoal cocoa seeds: {total}\nClass A: {count_classa} seeds\nClass B: {count_classb} seeds\nClass C: {count_classc} seeds"

    # Render results
    render = render_result(model=model, image=resized_image, result=results[0])

    print("Selected model:", selected_model)

    #return render, output_counts, plotbar
    return render, output_counts, "You have selected the " + str(selected_model)

#csspath = 'code/yolov8newultlt/gradio.css'

with gr.Blocks(theme='ParityError/LimeFace') as demo:

    with gr.Row():
        with gr.Column():
             gr.Interface(fn=detect_objects,
                          inputs=[gr.Image(type="filepath", label="Upload an Image"), gr.Dropdown(choices=["NST Model", "KCC Model"])],
                          outputs=[gr.Image(type="filepath", label="Result"), gr.Textbox(label="Detection Counts"), gr.Textbox(label="Selected Model")],
                          title="YOLOv8 Cocoa Seed Classification",
                          description="Upload an image to detect objects using YOLO.",
                          #html = gr.HTML(value="<p>This is another paragraph123.</p>"),
                          examples = [["NSTA.jpg"],
                                      ["NSTB.jpg"],
                                      ["NSTC.jpg"],
                                      ["KCCA.jpg"],
                                      ["KCCB.jpg"],
                                      ["KCCC.jpg"]],
                          cache_examples = bool(False)
                          #css=csspath,
                          )
             
    # with gr.Row(): #original Column
    #     with gr.Row(): #original Column
    #         example = [[gr.Image(value="NSTA.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')],
    #                    [gr.Image(value="NSTB.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')],
    #                    [gr.Image(value="NSTC.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')],
    #                    [gr.Image(value="KCCA.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')],
    #                    [gr.Image(value="KCCB.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')],
    #                    [gr.Image(value="KCCC.jpg", interactive = bool(True)), gr.Markdown(value='**label 5**')]]
    

    with gr.Row():
          with gr.Row():
            #   gr.HTML(value="<b>Class A</b> <p>Class A is the best from all 3 classes. It have the best of physical appreance eg. shape, size, texture</p>"),
            gr.HTML(value="<p>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; NSTA &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; NSTB &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; NTSC &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; KCCA &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; KCCB &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; KCCC</p> <dl> <dt><b>Class A</b></dt> <dd>Class A is the best from all 3 classes. It have the best of physical appreance eg. shape, size, texture</dd> </dl> <dt><b>Class B</b></dt> <dd>Class B most of the cocoa seed have physical appreance similar to class A. <br> But the size must me smaller and texture is not smmoth as class A</dd> <dt><b>Class C</b></dt> <dd>Class C is the worst from all 3 classes. Its the smallest, rough texter and have a irregular shape </dd> </dl></dl>")
         
if __name__ == "__main__":
    demo.queue().launch(share=True)