app.py

import cv2
import gradio as gr
import numpy as np
from fer import FER
from moviepy.editor import VideoFileClip


# Initialize the pre-trained detector once so you don't reinitialize on every function call.
detector = FER(mtcnn=True)  # Optionally, you can set mtcnn to False to use a faster (but less accurate) cascade.

def emotion_recognition(image):
    """
    Process the input image, detect emotions on faces,
    and annotate the image with bounding boxes and emotion labels.
    
    Parameters:
        image (numpy.ndarray): Input image (RGB).
    
    Returns:
        numpy.ndarray: Annotated image with emotion labels.
    """
    # fer works with RGB images which is what Gradio provides by default.
    results = detector.detect_emotions(image)
    annotated_image = image.copy()
    
    # Loop through each detected face
    for face in results:
        (x, y, w, h) = face["box"]
        # Get the dominant emotion for the detected face
        dominant_emotion = max(face["emotions"].items(), key=lambda item: item[1])[0]
        # Draw bounding box around face
        cv2.rectangle(annotated_image, (x, y), (x + w, y + h), (0, 255, 0), 2)
        # Put the emotion label above the bounding box
        cv2.putText(annotated_image, dominant_emotion, (x, y - 10), 
                    cv2.FONT_HERSHEY_SIMPLEX, 0.9, (36, 255, 12), 2)
    return annotated_image

# Create a Gradio Interface
interface = gr.Interface(
    fn=emotion_recognition,
    inputs=gr.Image(type="numpy", label="Input Image"),
    outputs=gr.Image(type="numpy", label="Annotated Image"),
    title="Facial Emotion Recognition",
    description="Upload an image and let the app detect and annotate facial emotions."
)

# Run the app locally
if __name__ == "__main__":
    interface.launch()