使用 YOLO11 分割和高斯模糊創建人像效果
作者：二旺
本文通過結合最新的 YOLO11 實例分割模型和高斯模糊，為你的圖片應用人像效果。
分割和高斯模糊后的圖像
本文通過結合最新的YOLO11實例分割模型和高斯模糊，為你的圖片應用人像效果。我們將使用YOLO11將人物從背景中分割出來，并對除了主體之外的所有內容應用模糊效果。
1. 安裝Ultralytics庫
首先創建并激活一個Python虛擬環境來管理依賴項。如果你不熟悉虛擬環境，請查看這個教程：
激活虛擬環境后，我們需要安裝ultralytics庫，這將允許我們使用YOLO11實例分割模型。運行以下命令在你的環境里安裝庫：
pip install ultralytics2. 下載測試圖片
接下來，讓我們從Unsplash下載一張測試圖片進行測試，你可以使用任何你選擇的圖片。我為我們的測試目的選擇了以下圖片：
在.py文件中，添加以下代碼來下載和加載圖片：
import urllib.request
import cv2
# Download the image
url, filename = ("https://images.unsplash.com/photo-1634646493821-9fca74f85f59?w=600&auto=format&fit=crop&q=60&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxzZWFyY2h8Mzk0fHx1bmJsdXJyZWQlMjBwb3J0YWl0fGVufDB8fDB8fHww", "scene.jpg")
urllib.request.urlretrieve(url, filename)
# Load the input image using OpenCV
image = cv2.imread(filename)3. 生成分割掩碼
圖片加載后，下一步是創建一個分割掩碼，以識別圖片中的人物。有關使用YOLO11實例分割模型識別人物的更詳細教程，請查看這個教程：《YOLO11 實例分割模型做行人分割》
模型將檢測人物，我們將創建一個掩碼以將主體與背景隔離。我們將使用yolo11n-seg.pt模型，但你可以使用Ultralytics YOLO11文檔中的任何你喜歡的模型。以下是加載模型并生成掩碼的代碼：
import urllib.request
import cv2
from ultralytics import YOLO
import numpy as np
def segment_image(image, model):
# Predict with the model
results = model(filename) # predict on an image
# Create an empty mask for segmentation
segmentation_mask = np.zeros_like(image, dtype=np.uint8)
# Iterate over the results
for i, r in enumerate(results):
# Iterate through the detected masks
for j, mask in enumerate(r.masks.xy):
# Convert the class tensor to an integer
class_id = int(r.boxes.cls[j].item()) # Extract the class ID as an integer
# Check if the detected class corresponds to 'person' (class ID 0)
if class_id == 0:
# Convert mask coordinates to an integer format for drawing
mask = np.array(mask, dtype=np.int32)
# Fill the segmentation mask with color (e.g., white for people)
cv2.fillPoly(segmentation_mask, [mask], (255, 255, 255))
return segmentation_mask
# Download the image
url, filename = ("https://images.unsplash.com/photo-1634646493821-9fca74f85f59?w=600&auto=format&fit=crop&q=60&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxzZWFyY2h8Mzk0fHx1bmJsdXJyZWQlMjBwb3J0YWl0fGVufDB8fDB8fHww", "scene.jpg")
urllib.request.urlretrieve(url, filename)
# Load the input image using OpenCV
image = cv2.imread(filename)
# Load the model
model = YOLO("yolo11n-seg.pt") # load an official YOLO model
# Generate the segmentation mask
segmentation_mask = segment_image(image, model)
# Visualize the segmentation mask before combining it with the original image
cv2.imwrite("mask.jpg", segmentation_mask)這一步將生成一個二進制掩碼，其中人物被突出顯示，如下例所示：
圖像二進制分割掩碼
4. 使用掩碼對圖像應用高斯模糊
現在我們有了分割掩碼，我們可以在保持人物清晰的同時對背景應用高斯模糊。我們將模糊整個圖像，然后使用掩碼將清晰的人物區域與模糊的背景結合起來。以下是分割和應用模糊的所有代碼：
import urllib.request
import cv2
from ultralytics import YOLO
import numpy as np
def segment_image(image, model):
# Predict with the model
results = model(filename) # predict on an image
# Create an empty mask for segmentation
segmentation_mask = np.zeros_like(image, dtype=np.uint8)
# Iterate over the results
for i, r in enumerate(results):
# Iterate through the detected masks
for j, mask in enumerate(r.masks.xy):
# Convert the class tensor to an integer
class_id = int(r.boxes.cls[j].item()) # Extract the class ID as an integer
# Check if the detected class corresponds to 'person' (class ID 0)
if class_id == 0:
# Convert mask coordinates to an integer format for drawing
mask = np.array(mask, dtype=np.int32)
# Fill the segmentation mask with color (e.g., white for people)
cv2.fillPoly(segmentation_mask, [mask], (255, 255, 255))
return segmentation_mask
def apply_blur_using_mask(image, mask, blur_strength=(25, 25)):
# Apply Gaussian blur to the entire image
blurred_image = cv2.GaussianBlur(image, blur_strength, 0)
# Create an inverted mask where the background is white and the person is black
inverted_mask = cv2.bitwise_not(mask)
# Use the mask to keep the person sharp and blur the background
background_blur = cv2.bitwise_and(blurred_image, blurred_image, mask=inverted_mask[:, :, 0])
person_region = cv2.bitwise_and(image, image, mask=mask[:, :, 0])
# Combine the sharp person region with the blurred background
final_image = cv2.add(person_region, background_blur)
return final_image
# Download the image
url, filename = ("https://images.unsplash.com/photo-1634646493821-9fca74f85f59?w=600&auto=format&fit=crop&q=60&ixlib=rb-4.0.3&ixid=M3wxMjA3fDB8MHxzZWFyY2h8Mzk0fHx1bmJsdXJyZWQlMjBwb3J0YWl0fGVufDB8fDB8fHww", "scene.jpg")
urllib.request.urlretrieve(url, filename)
# Load the input image using OpenCV
image = cv2.imread(filename)
# Load the model
model = YOLO("yolo11n-seg.pt") # load an official YOLO model
# Generate the segmentation mask
segmentation_mask = segment_image(image, model)
# Call the function to apply the blur and save the result
final_image = apply_blur_using_mask(image, segmentation_mask)
# Visualize the segmentation mask before combining it with the original image
cv2.imwrite("mask.jpg", segmentation_mask)
# Save the result
cv2.imwrite("blurred_image.jpg", final_image)
# Optionally display the image (make sure you're running in a GUI environment)
cv2.imshow("Blurred Image Result", final_image)
cv2.waitKey(0)
cv2.destroyAllWindows()最終結果
這段代碼將清晰的人物與模糊的背景結合起來，為你的圖像提供專業的人像效果。分割掩碼確保人物保持聚焦，而背景則通過高斯模糊變柔和。
示例結果
完整代碼：https://github.com/Brianhulela/background_blur
責任編輯：趙寧寧
來源：
小白玩轉Python
