Spaces:
Running
Running
import gradio as gr | |
from upstash_vector import AsyncIndex | |
from transformers import AutoFeatureExtractor, AutoModel | |
from datasets import load_dataset | |
index = AsyncIndex.from_env() | |
model_ckpt = "google/vit-base-patch16-224-in21k" | |
extractor = AutoFeatureExtractor.from_pretrained(model_ckpt) | |
model = AutoModel.from_pretrained(model_ckpt) | |
hidden_dim = model.config.hidden_size | |
dataset = load_dataset("BounharAbdelaziz/Face-Aging-Dataset") | |
MAX_K = 50 | |
with gr.Blocks() as demo: | |
gr.Markdown( | |
""" | |
# Find Your Twins | |
Upload your face and find the most similar faces from [Face Aging Dataset](https://huggingface.co/datasets/BounharAbdelaziz/Face-Aging-Dataset) using Google's [VIT](https://huggingface.co/google/vit-base-patch16-224-in21k) model. For best results please use 1x1 ratio face images, take a look at examples. Also increasing count in the advanced section results with more accurate searches. The Vector similarity search is powered by [Upstash Vector](https://upstash.com) 🚀. You can check our blog [post](https://huggingface.co/blog/omerXfaruq/serverless-image-similarity-with-upstash-vector) to learn more. | |
""" | |
) | |
with gr.Tab("Basic"): | |
with gr.Row(): | |
with gr.Column(scale=1): | |
input_image = gr.Image(type="pil") | |
with gr.Column(scale=2): | |
output_images = gr.Gallery() | |
async def find_similar_faces(image): | |
if image is None: | |
return None | |
inputs = extractor(images=image, return_tensors="pt") | |
outputs = model(**inputs) | |
embed = outputs.last_hidden_state[0][0] | |
result = await index.query(vector=embed.tolist(), top_k=4) | |
return [dataset["train"][int(vector.id)]["image"] for vector in result] | |
gr.Examples( | |
examples=[ | |
dataset["train"][6]["image"], | |
dataset["train"][7]["image"], | |
dataset["train"][8]["image"], | |
], | |
inputs=input_image, | |
outputs=output_images, | |
fn=find_similar_faces, | |
cache_examples=False, | |
) | |
with gr.Tab("Advanced"): | |
with gr.Row(): | |
with gr.Column(scale=1): | |
adv_input_image = gr.Image(type="pil") | |
adv_image_count = gr.Slider(1, MAX_K, 10, label="Image Count") | |
adv_button = gr.Button("Submit") | |
with gr.Column(scale=2): | |
adv_output_image = gr.Gallery() | |
async def find_similar_faces(image, count): | |
inputs = extractor(images=image, return_tensors="pt") | |
outputs = model(**inputs) | |
embed = outputs.last_hidden_state[0][0] | |
result = await index.query( | |
vector=embed.tolist(), top_k=max(1, min(MAX_K, int(count))) | |
) | |
return [dataset["train"][int(vector.id)]["image"] for vector in result] | |
adv_button.click( | |
fn=find_similar_faces, | |
inputs=[adv_input_image, adv_image_count], | |
outputs=[adv_output_image], | |
) | |
adv_input_image.upload( | |
fn=find_similar_faces, | |
inputs=[adv_input_image, adv_image_count], | |
outputs=[adv_output_image], | |
) | |
if __name__ == "__main__": | |
demo.queue(default_concurrency_limit=40) | |
demo.launch() | |