app.py

import gradio as gr
import torch
from PIL import Image
import numpy as np
import sys, os

# ---- Add YOLOv12 code to Python path ----
fork_ultra_path = os.path.join(os.path.dirname(__file__), "yolov12-main", "yolov12-main")
sys.path.insert(0, fork_ultra_path)

# ---- Import YOLO ----
from ultralytics import YOLO

# ---- Load trained model ----
model_path = os.path.join(os.path.dirname(__file__), "last.pt")
model = YOLO(model_path)

# ---- Detection function ----
def detect_objects(image):
    # Convert PIL to RGB NumPy array (YOLO expects RGB)
    img = np.array(image.convert("RGB"))

    # Run prediction with same settings used during training
    results = model.predict(
        source=img,
        imgsz=640,        # same resolution as training
        conf=0.25,        # lower confidence threshold to catch subtle detections
        iou=0.45,         # standard non-max suppression
        verbose=False
    )

    # Visualize results
    annotated = results[0].plot()
    return Image.fromarray(annotated)

# ---- Gradio Interface ----
demo = gr.Interface(
    fn=detect_objects,
    inputs=gr.Image(type="pil", label="Upload an Ear Image"),
    outputs=gr.Image(type="pil", label="Detection Result"),
    title="Ear Condition Detection",
    description="Upload an ear image to detect possible conditions using the trained YOLOv12 model."
)

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)