app.py

from transformers import AutoModelForImageClassification, AutoImageProcessor
import torch
import gradio as gr

model = AutoModelForImageClassification.from_pretrained("stnleyyg/visual-emotion-analyser", num_labels=8)
processor = AutoImageProcessor.from_pretrained("stnleyyg/visual-emotion-analyser")

def emotion_analyser(image):
    inputs = processor(image, return_tensors="pt")

    id2label = {
        "0": "anger",
        "1": "contempt",
        "2": "disgust",
        "3": "fear",
        "4": "happy",
        "5": "neutral",
        "6": "sad",
        "7": "surprise"
    }

    with torch.no_grad():
        logits = model(**inputs).logits
        probability = logits.softmax(-1)
        predicted_class_idx = probability.argmax(-1).item()

        predicted_label = id2label[str(predicted_class_idx)]

    return f"This person emotion is {predicted_label}"

demo_app = gr.Interface(
    fn=emotion_analyser,
    inputs=gr.Image(type="pil"),
    outputs=gr.TextArea()
)

demo_app.launch()