import requests

import gradio as gr
import torch
from timm import create_model
from timm.data import resolve_data_config
from timm.data.transforms_factory import create_transform

IMAGENET_1K_URL = 'https://storage.googleapis.com/bit_models/ilsvrc2012_wordnet_lemmas.txt'
LABELS = requests.get(IMAGENET_1K_URL).text.strip().split('\n')

model = create_model('resnet50', pretrained=True)

transform = create_transform(
    **resolve_data_config({}, model=model)
)
model.eval()

def predict_fn(img):
    img = img.convert('RGB')
    img = transform(img).unsqueeze(0)

    with torch.no_grad():
        out = model(img)
    
    probabilities = torch.nn.functional.softmax(out[0], dim=0)

    values, indices = torch.topk(probabilities, k=5)

    return {LABELS[i]: v.item() for i, v in zip(indices, values)}

gr.Interface(predict_fn, gr.inputs.Image(type='pil'), outputs='label').launch()