app.py

from PIL import Image
from transformers import BlipProcessor, BlipForConditionalGeneration
import torch
import cv2
import numpy as np
from deepface import DeepFace
import gradio as gr

# ====== 모델 로딩 ======
processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

# ====== 옷 정보 추출 함수 ======
def extract_clothing(text):
    colors = ['red', 'blue', 'green', 'black', 'white', 'yellow', 'brown', 'gray', 'pink', 'orange']
    patterns = ['striped', 'checkered', 'plaid', 'polka-dot', 'solid', 'patterned', 'floral']
    items = ['jacket', 'coat', 'dress', 'shirt', 't-shirt', 'jeans', 'pants', 'shorts',
             'suit', 'sneakers', 'hat', 'scarf', 'uniform']
    
    found_colors = [c for c in colors if c in text.lower()]
    found_patterns = [p for p in patterns if p in text.lower()]
    found_items = [i for i in text.lower().split() if i in items]
    
    return found_colors, found_patterns, found_items

# ====== 최종 설명 생성 함수 ======
def generate_15_sentences(caption, num_faces, age_summary, clothing_sentence):
    sentences = []
    sentences.append(f"The image presents the scene: {caption}.")
    sentences.append("The visual tone combines human presence with context-rich elements.")
    sentences.append(f"A total of {num_faces} people with visible faces were detected.")
    
    if age_summary:
        summary_list = [f"{v} {k}(s)" for k, v in age_summary.items()]
        sentences.append("The crowd includes " + ", ".join(summary_list) + ".")
    else:
        sentences.append("No specific age or gender details were identified.")
    
    sentences.append(clothing_sentence)
    sentences.append("Facial expressions range from neutral to slightly expressive, adding emotional context.")
    sentences.append("Some individuals appear to be interacting with the environment or each other.")
    sentences.append("Although specific facial shapes are not automatically classified here, a mix of face sizes and angles is present.")
    sentences.append("Hairstyles vary, including short hair, longer cuts, and tied-back styles depending on individual orientation.")
    sentences.append("The photo captures diversity not only in people but also in visual textures and tones.")
    sentences.append("Clothing styles vary, suggesting informal or casual settings rather than formal events.")
    sentences.append("The spatial arrangement of individuals indicates natural movement or candid posture.")
    sentences.append("Background elements such as buildings or trees provide additional narrative depth.")
    sentences.append("The lighting helps highlight human features and adds dimensionality to the scene.")
    sentences.append("Overall, the image blends appearance, age, fashion, and emotion into a coherent story.")
    return sentences

# ====== 메인 분석 함수 ======
def analyze_uploaded_image(image_pil):
    image_pil = image_pil.convert("RGB")
    image_np = np.array(image_pil)

    # 1. Caption 생성 (BLIP)
    inputs = processor(image_pil, return_tensors="pt")
    out = model.generate(**inputs)
    caption = processor.decode(out[0], skip_special_tokens=True)

    # 2. 얼굴 감지 (OpenCV)
    gray = cv2.cvtColor(image_np, cv2.COLOR_RGB2GRAY)
    faces = face_cascade.detectMultiScale(gray, 1.1, 4)

    # 3. DeepFace로 연령/성별 분석
    face_infos = []
    for (x, y, w, h) in faces:
        face_crop = image_np[y:y+h, x:x+w]
        try:
            analysis = DeepFace.analyze(face_crop, actions=['age', 'gender'], enforce_detection=False)
            age = analysis[0]['age']
            gender = analysis[0]['gender']
            if age < 13:
                age_group = "child"
            elif age < 20:
                age_group = "teen"
            elif age < 60:
                age_group = "adult"
            else:
                age_group = "senior"
            face_infos.append({
                "age_group": age_group,
                "gender": gender,
            })
        except:
            continue

    num_faces = len(face_infos)

    # 4. 연령대 요약
    age_summary = {}
    for face in face_infos:
        key = f"{face['gender']} {face['age_group']}"
        age_summary[key] = age_summary.get(key, 0) + 1

    # 5. 의복 정보 추출
    colors, patterns, items = extract_clothing(caption)
    parts = []
    if colors:
        parts.append(f"colors such as {', '.join(colors)}")
    if patterns:
        parts.append(f"patterns like {', '.join(patterns)}")
    if items:
        parts.append(f"clothing items such as {', '.join(items)}")
    clothing_sentence = "The clothing observed includes " + " with ".join(parts) + "." if parts else "Clothing is present but not clearly distinguishable."

    # 6. 최종 설명 생성
    final_description = generate_15_sentences(caption, num_faces, age_summary, clothing_sentence)
    return "\n".join([f"{i+1}. {s}" for i, s in enumerate(final_description)])

# ====== Gradio 인터페이스 설정 ======
interface = gr.Interface(
    fn=analyze_uploaded_image,
    inputs=gr.Image(type="pil", label="이미지를 업로드하세요"),
    outputs=gr.Textbox(label="15문장 이미지 설명"),
    title="🧠 이미지 인식 설명기 (BLIP + DeepFace)",
    description="이미지를 업로드하면 사람 수, 성별, 연령대, 옷, 분위기 등을 15개의 문장으로 설명합니다."
)

# ====== 앱 실행 ======
interface.launch()