In [2]:
# Sample Image
with Image.open("./Data/images/maksssksksss0.png") as rawImage:
display(rawImage)
# Import Packages
import os
import re
import tensorflow as tf
import pandas as pd
from bs4 import BeautifulSoup as soup
from PIL import Image
import matplotlib.pyplot as plot
import matplotlib.patches as patches
# Sample Image
with Image.open("./Data/images/maksssksksss0.png") as rawImage:
display(rawImage)
# Sample Annotation
with open("./Data/annotations/maksssksksss0.xml") as annotation:
rawXML = annotation.read()
print(rawXML)
<annotation>
<folder>images</folder>
<filename>maksssksksss0.png</filename>
<size>
<width>512</width>
<height>366</height>
<depth>3</depth>
</size>
<segmented>0</segmented>
<object>
<name>without_mask</name>
<pose>Unspecified</pose>
<truncated>0</truncated>
<occluded>0</occluded>
<difficult>0</difficult>
<bndbox>
<xmin>79</xmin>
<ymin>105</ymin>
<xmax>109</xmax>
<ymax>142</ymax>
</bndbox>
</object>
<object>
<name>with_mask</name>
<pose>Unspecified</pose>
<truncated>0</truncated>
<occluded>0</occluded>
<difficult>0</difficult>
<bndbox>
<xmin>185</xmin>
<ymin>100</ymin>
<xmax>226</xmax>
<ymax>144</ymax>
</bndbox>
</object>
<object>
<name>without_mask</name>
<pose>Unspecified</pose>
<truncated>0</truncated>
<occluded>0</occluded>
<difficult>0</difficult>
<bndbox>
<xmin>325</xmin>
<ymin>90</ymin>
<xmax>360</xmax>
<ymax>141</ymax>
</bndbox>
</object>
</annotation>
# Data Extraction Function
def Extract(path):
# Safely open annotation
with open(path) as annotation:
# Read annotation with Beautiful Soup
rawXML = annotation.read()
bsXML = soup(rawXML,"xml")
# Stores Metadata in a Dictionary
meta = {}
size = bsXML.find('size')
meta['index']=int(re.search(r'\d+',path).group())
meta['width'] = int(size.find('width').get_text())
meta['height'] = int(size.find('height').get_text())
meta['depth'] = int(size.find('depth').get_text())
# Stores Labels in a List of Dictionaries
faces = []
objects = bsXML.find_all('object')
for face in objects:
labels={}
labels['index']=meta['index']
labels['label'] = face.find('name').get_text()
labels['pose'] = face.find('pose').get_text()
labels['truncated'] = int(face.find('truncated').get_text())
labels['occluded'] = int(face.find('occluded').get_text())
labels['difficult'] = int(face.find('difficult').get_text())
labels['xmin'] = int(face.find('xmin').get_text())
labels['xmax'] = int(face.find('xmax').get_text())
labels['ymin'] = int(face.find('ymin').get_text())
labels['ymax'] = int(face.find('ymax').get_text())
faces.append(labels)
return(meta,faces)
# Extract data from sample
meta,faces = Extract("./Data/annotations/maksssksksss0.xml")
print("Metadata: ")
print(meta)
print("Face Objects: ")
print(faces)
Metadata:
{'index': 0, 'width': 512, 'height': 366, 'depth': 3}
Face Objects:
[{'index': 0, 'label': 'without_mask', 'pose': 'Unspecified', 'truncated': 0, 'occluded': 0, 'difficult': 0, 'xmin': 79, 'xmax': 109, 'ymin': 105, 'ymax': 142}, {'index': 0, 'label': 'with_mask', 'pose': 'Unspecified', 'truncated': 0, 'occluded': 0, 'difficult': 0, 'xmin': 185, 'xmax': 226, 'ymin': 100, 'ymax': 144}, {'index': 0, 'label': 'without_mask', 'pose': 'Unspecified', 'truncated': 0, 'occluded': 0, 'difficult': 0, 'xmin': 325, 'xmax': 360, 'ymin': 90, 'ymax': 141}]
# Show labels on Image
with Image.open("./Data/images/maksssksksss0.png") as rawImage:
figure, axis = plot.subplots()
axis.imshow(rawImage)
meta,objects = Extract("./Data/annotations/maksssksksss0.xml")
for face in objects:
color = 'green' if face['label']=='with_mask' else 'red'
corner = (face['xmin'],face['ymin'])
width = face['xmax']-face['xmin']
height = face['ymax']-face['ymin']
rect = patches.Rectangle(corner, width, height, linewidth=2, edgecolor=color, facecolor='none')
axis.add_patch(rect)
plot.show()
# Collect Samples
imagePath = "./Data/images/"
imagesList = [[int(re.search(r'\d+',os.path.join(imagePath,f)).group()),os.path.join(imagePath,f)] for f in os.listdir(imagePath) if os.path.isfile(os.path.join(imagePath,f))]
images = pd.DataFrame(imagesList,columns=['index','path']).set_index('index')
# Collect Labels
annotationpath = "./Data/annotations/"
metaList=[]
facesList=[]
for f in os.listdir(annotationpath):
path = os.path.join(annotationpath,f)
meta, faces = Extract(path)
metaList.append(meta)
facesList=facesList+faces
meta = pd.DataFrame(metaList).set_index('index')
labels = pd.DataFrame(facesList).set_index('index')
# Metadata Summary Statistics
meta.describe()
| width | height | depth | |
|---|---|---|---|
| count | 853.000000 | 853.000000 | 853.0 |
| mean | 370.589683 | 309.289566 | 3.0 |
| std | 56.224049 | 74.474840 | 0.0 |
| min | 193.000000 | 156.000000 | 3.0 |
| 25% | 301.000000 | 254.000000 | 3.0 |
| 50% | 400.000000 | 280.000000 | 3.0 |
| 75% | 400.000000 | 400.000000 | 3.0 |
| max | 600.000000 | 600.000000 | 3.0 |
# Explore Image Sizes
groupedMeta = meta.groupby(['height', 'width']).size().reset_index().rename(columns={0:'count'})
print("Most images of the same size:")
print(groupedMeta[groupedMeta['count']==groupedMeta['count'].max()])
plot.scatter(groupedMeta['width'],groupedMeta['height'],c='darkblue',s=3*groupedMeta['count'])
plot.xlabel('width (pixels)')
plot.ylabel('height (pixels)')
plot.title('Sizes of Images in Dataset')
plot.show()
Most images of the same size:
height width count
115 400 301 184
# Labels Summary Statistics
nlabels = len(labels)
nWithMask = len(labels[labels['label']=='with_mask'])
nWithoutMask = len(labels[labels['label']=='without_mask'])
nIncorrectMask = len(labels[labels['label']=='mask_weared_incorrect'])
nOther = nlabels - nWithMask - nWithoutMask - nIncorrectMask
labelTable = pd.DataFrame([
["With","{:.1f}".format(100*nWithMask/nlabels)+'%'],
["Without","{:.1f}".format(100*nWithoutMask/nlabels)+"%"],
["Incorrect","{:.1f}".format(100*nIncorrectMask/nlabels)+"%"]],columns=["Mask Label","Label Classifications ("+str(nlabels)+" Total Labels)"]).set_index("Mask Label")
labelTable
| Label Classifications (4072 Total Labels) | |
|---|---|
| Mask Label | |
| With | 79.4% |
| Without | 17.6% |
| Incorrect | 3.0% |
# Explore Number of Labels per Image
groupedLabels = labels.groupby(by='index').size()
plot.hist(groupedLabels,bins=25,color='teal')
plot.xlabel('Labels in Image')
plot.ylabel('Number of Images')
plot.title('Distribution of Labels in Images')
plot.show()
# Look into other label variables
print(labels.groupby(by='pose').size())
labels.describe()
pose Unspecified 4072 dtype: int64
| truncated | occluded | difficult | xmin | xmax | ymin | ymax | |
|---|---|---|---|---|---|---|---|
| count | 4072.0 | 4072.0 | 4072.0 | 4072.000000 | 4072.000000 | 4072.000000 | 4072.000000 |
| mean | 0.0 | 0.0 | 0.0 | 182.207024 | 213.356090 | 85.780697 | 120.785609 |
| std | 0.0 | 0.0 | 0.0 | 104.471254 | 102.712267 | 52.571821 | 70.355098 |
| min | 0.0 | 0.0 | 0.0 | 1.000000 | 8.000000 | 1.000000 | 6.000000 |
| 25% | 0.0 | 0.0 | 0.0 | 96.000000 | 134.000000 | 49.000000 | 73.000000 |
| 50% | 0.0 | 0.0 | 0.0 | 177.000000 | 212.000000 | 75.000000 | 103.000000 |
| 75% | 0.0 | 0.0 | 0.0 | 266.000000 | 292.000000 | 113.000000 | 148.000000 |
| max | 0.0 | 0.0 | 0.0 | 569.000000 | 592.000000 | 330.000000 | 495.000000 |