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") TOP_K = 1000 BASE_COUNT=4 MAX_COUNT = 10 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. Disclaimer, this demo doesn't find your twins :), it finds similar face parts, shapes, features(nose, cheek, face, forehead shapes) that are encoded in the model. 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() @input_image.change(inputs=input_image, outputs=output_images) 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=TOP_K) return [dataset["train"][int(vector.id)]["image"] for vector in result[:BASE_COUNT]] 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_COUNT, BASE_COUNT, label="Image Count") adv_button = gr.Button("Submit") with gr.Column(scale=2): adv_output_images = gr.Gallery() async def find_similar_faces(image, count): 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=TOP_K ) return [dataset["train"][int(vector.id)]["image"] for vector in result[:int(count)]] adv_button.click( fn=find_similar_faces, inputs=[adv_input_image, adv_image_count], outputs=[adv_output_images], ) adv_input_image.change( fn=find_similar_faces, inputs=[adv_input_image, adv_image_count], outputs=[adv_output_images], ) gr.Examples( examples=[ [dataset["train"][6]["image"], BASE_COUNT], [dataset["train"][7]["image"], BASE_COUNT], [dataset["train"][8]["image"], BASE_COUNT], ], inputs=[adv_input_image, adv_image_count], outputs=adv_output_images, fn=find_similar_faces, cache_examples=False, ) if __name__ == "__main__": demo.queue(default_concurrency_limit=40) demo.launch()