from transformers import pipeline from PIL import Image, ImageDraw from sentence_transformers import SentenceTransformer, util import requests from io import BytesIO import gradio as gr # Inisialisasi model deteksi dan pengkodean checkpoint = "google/owlvit-base-patch32" detector = pipeline(model=checkpoint, task="zero-shot-object-detection") model = SentenceTransformer('clip-ViT-L-14') def load_image(url): response = requests.get(url) image = Image.open(BytesIO(response.content)).convert("RGB") return image def get_face_image(url): image = load_image(url) predictions = detector( image, candidate_labels=["human face"], ) max_score = 0 box_area = None for prediction in predictions: box = prediction["box"] label = prediction["label"] score = prediction["score"] if score > max_score: xmin, ymin, xmax, ymax = box.values() box_area = (xmin, ymin, xmax, ymax) max_score = score else: continue if box_area: draw = ImageDraw.Draw(image) draw.rectangle(box_area, outline="red", width=2) crop_img = image.crop(box_area) newsize = (256, 256) face_img = crop_img.resize(newsize) return face_img def predict(url1, url2, inp_sim): face_image1 = get_face_image(url1) face_image2 = get_face_image(url2) if face_image1 is not None and face_image2 is not None: img_emb = model.encode([face_image1, face_image2]) sim = util.cos_sim(img_emb[0], img_emb[1]) if sim > inp_sim: return sim.item(), "SAME" else: return sim.item(), "DIFFERENT" else: return 0, "No face detected in one or both images" description = "An application that can recognize if two faces belong to the same person or not" title = "Facial Identity Recognition System" interface = gr.Interface(fn=predict, inputs=[gr.Textbox(label="Image URL 1"), gr.Textbox(label="Image URL 2"), gr.Slider(0, 1, value=0.8, label="Similarity Percentage", info="Choose between 0 and 1")], outputs=[gr.Number(label="Similarity"), gr.Textbox(label="Message")], description=description, title=title) interface.launch(debug=True)