利用PaddleHub戴绿帽[人间的青草地需要浇水~]
被抖音洗脑的人间的青草地需要浇水~
驱使了我想要做本项目,项目已放在AIstudio
话不多说，开干！
# 安装Paddlehub !pip install paddlehub==2.1.0
! hub install ultra_light_fast_generic_face_detector_1mb_640==1.1.2
通过paddlehub获取位置信息
import paddlehub as hub module = hub.Module(name="ultra_light_fast_generic_face_detector_1mb_640", version="1.1.2") res = module.face_detection( paths = ["data/data96296/1.jpg"], visualization=True, output_dir='face_detection_output')
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/__init__.py:107: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import MutableMapping /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/rcsetup.py:20: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import Iterable, Mapping /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/colors.py:53: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working from collections import Sized [2021-06-22 16:33:08,556] [ WARNING] - The _initialize method in HubModule will soon be deprecated, you can use the __init__() to handle the initialization of the object
环境准备
import cv2 from PIL import Image import numpy as np from matplotlib import pyplot as plt %matplotlib inline filename = 'face_detection_output/1.jpg' ## [Load an image from a file] img = cv2.imread(filename) img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) plt.imshow(img)
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2349: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working if isinstance(obj, collections.Iterator): /opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/matplotlib/cbook/__init__.py:2366: DeprecationWarning: Using or importing the ABCs from 'collections' instead of from 'collections.abc' is deprecated, and in 3.8 it will stop working return list(data) if isinstance(data, collections.MappingView) else data <matplotlib.image.AxesImage at 0x7fe05c0632d0>
res # 返回 # res (list[dict]): 识别结果的列表，列表中每一个元素为 dict，关键字有'path', 'data'，相应的取值为： # path (str): 原输入图片的路径； # data (numpy.ndarray): 图像分割得到的结果，shape 为H * W，元素的取值为0-19，表示每个像素的分类结果，映射顺序与下面的调色板相同。
[{'data': [{'left': 221.29409790039062, 'right': 488.7735290527344, 'top': 78.7082290649414, 'bottom': 471.13763427734375, 'confidence': 0.9999988079071045}], 'path': 'data/data96296/1.jpg', 'save_path': 'face_detection_output/1.jpg'}]
# 处理为整数，小数不符合后续位置参数的要求 import math data = res[0]['data'] left = data[0]['left'] left = math.ceil(left) right = data[0]['right'] right = math.ceil(right) top = data[0]['top'] top = math.ceil(top) bottom = data[0]['bottom'] bottom = math.ceil(bottom) print("left:",left) print("right:",right) print("top:",top) print("bottom:",bottom)
left: 222 right: 489 top: 79 bottom: 472
有上面的数据可以知道:
框的左上角顶点位置为[left,top]
框的右上角顶点位置为[right,top]
得到这些信息后，再带上提前做好的帽子，就思路清晰了
将帽子尺寸改为[right-left+25,2.2*top] 25和2.2可以自己进行调整，合适为止。
# 储存帽子数据 hat_w , hat_h= right-left+40, 2.2*top hat_w = math.ceil(hat_w) hat_h = math.ceil(hat_h) print(hat_w) print(hat_h)
307 174
图片合成
from PIL import Image mother_img = "data/data96296/1.jpg" son_img = "data/data96296/lmz.png" save_img = "result/result.png" # 获取图片,方便后面的代码调用 M_Img = Image.open(mother_img) S_Img = Image.open(son_img) # 子图缩小的倍数1代表不变，2就代表原来的一半（没啥用，本来想通过这个控制图片） factor = 1 # 给图片指定色彩显示格式 S_Img = S_Img.convert("RGBA") # 获取图片的尺寸 M_Img_w, M_Img_h = M_Img.size # 获取被放图片的大小（母图） print("母图尺寸：",M_Img.size) S_Img_w, S_Img_h = S_Img.size # 获取小图的大小（子图） print("子图尺寸：",S_Img.size) size_w = int(S_Img_w / factor) size_h = int(S_Img_h / factor) # 防止子图尺寸大于母图 if S_Img_w > size_w: S_Img_w = size_w if S_Img_h > size_h: S_Img_h = size_h # 重新设置子图的尺寸 icon = S_Img.resize((hat_w, hat_h), Image.ANTIALIAS) # 子图的左上角坐标位置 w = left-20 h = 0 # 粘贴子图到母图的指定坐标（当前居中） M_Img.paste(icon, (w, h), mask=None) # 保存图片 M_Img.save(save_img)
母图尺寸： (693, 539) 子图尺寸： (224, 236)
查看效果
filename = 'result/result.png' ## [Load an image from a file] img = cv2.imread(filename) img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) plt.imshow(img)
<matplotlib.image.AxesImage at 0x7fe050429050>
总结
- 利用PaddleHub获取人体的头部位置
- 合理计算出帽子所需要放置的位置
- 图片合成
- 很明显（我这个版本很失败，还没有找到好的办法去把图片的黑色去掉，郁闷（明明帽子图片是PNG且代码也写了RGBA））
