Vibecode to blur multiple people using AI in Python

blur_multiple_people.py

import cv2
import numpy as np
from ultralytics import YOLO

# --- Configuration ---
DEBUG = True
IMAGE_PATH = "image.jpeg"
OUTPUT_PATH = (
    f"{'debug-' if DEBUG else ''}blurred-image.jpeg"
)
YOLO_MODEL_PATH = "yolo12x.pt"
FACE_PROTO_PATH = "deploy.prototxt"
FACE_MODEL_PATH = "res10_300x300_ssd_iter_140000.caffemodel"

SCALES = [0.5, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.5]  # Try more if needed
YOLO_CONF = 0.005  # Lower confidence catches more, but may increase false positives
YOLO_IOU = 0.05  # Lower IOU allows more overlapping boxes
YOLO_NMS = 0.99  # Higher NMS keeps more boxes
FACE_CONF = 0.15
FACE_PERSON_IOU = 0.1  # Allowed IOU for face/person overlap


def box_iou(boxA, boxB):
    xA = max(boxA[0], boxB[0])
    yA = max(boxA[1], boxB[1])
    xB = min(boxA[2], boxB[2])
    yB = min(boxA[3], boxB[3])
    interArea = max(0, xB - xA) * max(0, yB - yA)
    boxAArea = (boxA[2] - boxA[0]) * (boxA[3] - boxA[1])
    boxBArea = (boxB[2] - boxB[0]) * (boxB[3] - boxB[1])
    if boxAArea + boxBArea - interArea == 0:
        return 0
    return interArea / float(boxAArea + boxBArea - interArea)


# --- Load image ---
image = cv2.imread(IMAGE_PATH)
if image is None:
    raise IOError(f"Cannot load image: {IMAGE_PATH}")
h, w = image.shape[:2]

# --- Load YOLO model ---
model = YOLO(YOLO_MODEL_PATH)

# --- Multi-scale YOLO person detection ---
person_boxes = []
person_scores = []

for scale in SCALES:
    if DEBUG:
        print(f"Scale {scale}:")
    scaled_img = cv2.resize(image, (int(w * scale), int(h * scale)))
    results = model(scaled_img, conf=YOLO_CONF, iou=YOLO_IOU)
    for result in results:
        for box in result.boxes:
            cls_id = int(box.cls[0])
            if model.names[cls_id].lower() == "person":
                x1, y1, x2, y2 = map(int, box.xyxy[0])
                # Scale coordinates back to original image
                x1 = int(x1 / scale)
                y1 = int(y1 / scale)
                x2 = int(x2 / scale)
                y2 = int(y2 / scale)
                conf = float(box.conf[0])
                person_boxes.append([x1, y1, x2, y2])
                person_scores.append(conf)
    if DEBUG:
        print(f"  Persons found: {len(person_boxes)}")

# --- Apply Non-Maximum Suppression (NMS) for person boxes ---
indices = cv2.dnn.NMSBoxes(
    person_boxes, person_scores, score_threshold=YOLO_CONF, nms_threshold=YOLO_NMS
)
if len(indices) > 0:
    indices = np.array(indices).flatten()
    final_person_boxes = [person_boxes[i] for i in indices]
else:
    final_person_boxes = []
if DEBUG:
    print(f"After NMS: {len(final_person_boxes)} person boxes.")

# --- Face detection using OpenCV DNN ---
face_net = cv2.dnn.readNetFromCaffe(FACE_PROTO_PATH, FACE_MODEL_PATH)
blob = cv2.dnn.blobFromImage(
    cv2.resize(image, (300, 300)), 1.0, (300, 300), (104.0, 177.0, 123.0)
)
face_net.setInput(blob)
detections = face_net.forward()

print("Detections shape:", detections.shape)
for i in range(detections.shape[2]):
    confidence = detections[0, 0, i, 2]
    print(f"Detection {i}: confidence={confidence}")

face_boxes = []
for i in range(detections.shape[2]):
    confidence = detections[0, 0, i, 2]
    if confidence > FACE_CONF:
        box = detections[0, 0, i, 3:7] * np.array([w, h, w, h])
        x1, y1, x2, y2 = box.astype("int")
        x1, y1, x2, y2 = max(x1, 0), max(y1, 0), min(x2, w), min(y2, h)
        # Only add if not significantly overlapping a detected person box
        is_inside = False
        for px1, py1, px2, py2 in final_person_boxes:
            if box_iou([x1, y1, x2, y2], [px1, py1, px2, py2]) > FACE_PERSON_IOU:
                is_inside = True
                break
        if not is_inside:
            face_boxes.append([x1, y1, x2, y2])
if DEBUG:
    print(f"Face boxes (outside person): {len(face_boxes)}")

# --- Draw rectangles on image ---
output_img = image.copy()
for x1, y1, x2, y2 in final_person_boxes:
    person = image[y1:y2, x1:x2]
    if person.size > 0:
        if DEBUG:
            cv2.rectangle(
                output_img, (x1, y1), (x2, y2), (0, 255, 0), 3
            )  # Green rectangle with thickness 3 for people
        blurred = cv2.GaussianBlur(person, (51, 51), 0)
        output_img[y1:y2, x1:x2] = blurred
for x1, y1, x2, y2 in face_boxes:
    face = image[y1:y2, x1:x2]
    if face.size > 0:
        if DEBUG:
            cv2.rectangle(
                output_img, (x1, y1), (x2, y2), (255, 0, 0), 2
            )  # Blue for faces
        blurred = cv2.GaussianBlur(face, (51, 51), 0)
        output_img[y1:y2, x1:x2] = blurred
# --- Save output ---
cv2.imwrite(OUTPUT_PATH, output_img)
print(f"Output saved as {OUTPUT_PATH}")
print(
    f"People detected: {len(final_person_boxes)} | Faces detected (outside person boxes): {len(face_boxes)}"
)