import gradio as gr
from transformers import FlavaModel, BertTokenizer, FlavaFeatureExtractor
import numpy as np
from PIL import Image
import torch


images="dog.jpg"
model = FlavaModel.from_pretrained("facebook/flava-full")
model.eval()
fe = FlavaFeatureExtractor.from_pretrained("facebook/flava-full")
tokenizer = BertTokenizer.from_pretrained("facebook/flava-full")

def shot(image, labels_text):
    PIL_image = Image.fromarray(np.uint8(image)).convert('RGB')
    labels = labels_text.split(",")
    label_with_template = [f"This is a photo of a {label}" for label in labels]
    image_input = fe([PIL_image], return_tensors="pt")
    text_inputs = tokenizer(label_with_template, padding="max_length", return_tensors="pt")

    image_embeddings = model.get_image_features(**image_input)[:, 0, :]
    text_embeddings = model.get_text_features(**text_inputs)[:, 0, :]
    similarities = list(torch.nn.functional.softmax((text_embeddings @ image_embeddings.T).squeeze(0), dim=0))
    return {label: similarities[idx].item() for idx, label in enumerate(labels)}
    
iface = gr.Interface(shot, 
                    ["image", "text"], 
                    "label", 
                    examples=[["dog.jpg", "dog,cat,bird"], 
                              ["germany.jpg", "germany,belgium,colombia"],
                              ["rocket.jpg", "car,rocket,train"]
                            ],
                    description="Add a picture and a list of labels separated by commas",
                    title="FLAVA Zero-shot Image Classification")

iface.launch()