app.py

import gradio as gr
import torch
from ultralytics import YOLO
import os

REPO_URL = "https://github.com/WildHackers/community-fish-detector"
MODEL_URL = REPO_URL + "/releases/download/cfd-1.00-yolov12x/cfd-yolov12x-1.00.pt"

# Download model once
MODEL_PATH = os.path.basename(MODEL_URL)
if not os.path.exists(MODEL_PATH):
    torch.hub.download_url_to_file(MODEL_URL, MODEL_PATH)

# Load YOLOv12x model
model = YOLO(MODEL_PATH)

def run_detection(input_image, conf_threshold: float = 0.60, iou_threshold: float = 0.45, imgsz: int = 1024):
    """
    Runs YOLOv12x inference on an image.
    Returns annotated image result.
    """
    if input_image is None:
        return None

    results = model.predict(
        source=input_image,
        conf=conf_threshold,
        iou=iou_threshold,
        imgsz=imgsz,
        save=False,
        verbose=False
    )

    return results[0].plot()

# Gradio interface
demo = gr.Interface(
    fn=run_detection,
    inputs=[
        gr.Image(type="numpy", label="Input Image"),
        gr.Slider(0, 1, value=0.60, step=0.01, label="Confidence Threshold"),
        gr.Slider(0, 1, value=0.45, step=0.01, label="IoU Threshold"),
        gr.Slider(320, 1280, value=1024, step=32, label="Image Size"),
    ],
    outputs=gr.Image(type="numpy", label="Detected Output"),
    title="Community Fish Detector (YOLOv12x)",
    description=(
        f"Upload an image to detect fish using the [Community Fish Detector]({REPO_URL})."
    ),
    flagging_mode="never",
)

if __name__ == "__main__":
    demo.launch()