app.py

import torch
import os
import gradio as gr

from model import create_vit
from timeit import default_timer as timer
from typing import Tuple, Dict

class_names = ['cardboard', 'glass', 'metal', 'paper', 'plastic', 'trash']

vit, vit_transform = create_vit(output_classes=len(class_names))

vit.load_state_dict(torch.load(f="vit_b_16_dout0.3_10epochs.pth", map_location=torch.device("cpu")))

def predict(img) -> Tuple[Dict, float]:
    start_time = timer()
    img = vit_transform(img).unsqueeze(0)

    vit.eval()
    with torch.inference_mode():
        pred_probs = torch.softmax(vit(img), dim=1)
    
    pred_labels_and_probs = {class_names[i]: float(pred_probs[0][i]) for i in range(len(class_names))}
    pred_time = round(timer() - start_time, 5)
    return pred_labels_and_probs, pred_time

title = "Garbage Sense"
description = "A vision transformer trained to classify garbage into 6 categories on [trashnet](https://github.com/garythung/trashnet)."
article = ""

example_list = [["examples/" + example] for example in os.listdir("examples")]

demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs=[
        gr.Label(num_top_classes=6, label="Predictions"),
        gr.Number(label="Prediction time (s)"),
    ],
    examples=example_list,
    title=title,
    description=description
)

demo.launch()