app.py

import gradio as gr
from huggingface_hub import hf_hub_download
import onnxruntime
from huggingface_hub import ModelCard

card = ModelCard.load('mkhug98/Echo-Yolo')
# Download the ONNX model from Hugging Face
model_path = hf_hub_download(repo_id="mkhug98/Echo-Yolo", filename="best.onnx")

# Load the ONNX model
session = onnxruntime.InferenceSession(model_path)

# Function to perform object detection
def detect_objects(image):
    # Preprocess the image
    image = image.resize((640, 640))  # Resize the image to the expected input size
    input_data = image.transpose(2, 0, 1).numpy()  # Rearrange the dimensions for ONNX

    # Perform inference with the ONNX model
    outputs = session.run(None, {"images": input_data.astype("float32")})
    bboxes, scores, class_ids = outputs

    # Create a list of dictionaries for each detected object
    detections = []
    for bbox, score, class_id in zip(bboxes[0], scores[0], class_ids[0]):
        x1, y1, x2, y2 = bbox
        label = session.get_modelmeta().custom_metadata_map["names"][int(class_id)]
        detections.append({
            'label': label,
            'confidence': float(score),
            'x1': float(x1),
            'y1': float(y1),
            'x2': float(x2),
            'y2': float(y2)
        })

    return detections

# Create the Gradio app
app = gr.Interface(detect_objects, gr.Image(type="pil"), "label", examples=[
    ["example_image.jpg"]  # Replace with your own example image
])

# Run the app
app.launch()