DataHandler.py

'''
Original script : https://www.kaggle.com/lplenka/coco-data-visualization
Installing pycocotools : pip install 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'
'''


from pycocotools.coco import COCO
import numpy as np
import skimage.io as io
import matplotlib.pyplot as plt
import pylab
import random
pylab.rcParams['figure.figsize'] = (8.0, 10.0)# Import Libraries


dataDir='./val'
dataType='COCO_val_annos'
mul_dataType='COCO_mul_val_annos'
annFile='{}/{}.json'.format(dataDir,dataType)
mul_annFile='{}/{}.json'.format(dataDir,mul_dataType)
img_dir = "./img"

# initialize coco api for instance annotations
coco=COCO(annFile)
mul_coco=COCO(mul_annFile)

# display categories and supercategories

#Single Class #Damage dataset
cats = coco.loadCats(coco.getCatIds())
nms=[cat['name'] for cat in cats]
print('COCO categories for damages: \n{}\n'.format(', '.join(nms)))

nms = set([cat['supercategory'] for cat in cats])
print('COCO supercategories for damages: \n{}\n'.format(', '.join(nms)))

#Multi Class #Parts dataset

mul_cats = mul_coco.loadCats(mul_coco.getCatIds())
mul_nms=[cat['name'] for cat in mul_cats]
print('COCO categories for parts: \n{}\n'.format(', '.join(mul_nms)))

mul_nms = set([mul_cat['supercategory'] for mul_cat in mul_cats])
print('COCO supercategories for parts: \n{}\n'.format(', '.join(mul_nms)))

# get all images containing 'damage' category, select one at random
catIds = coco.getCatIds(catNms=['damage'])
imgIds = coco.getImgIds(catIds=catIds )

random_img_id = random.choice(imgIds)
print("{} image id was selected at random from the {} list".format(random_img_id, imgIds))

# Load the image
imgId = coco.getImgIds(imgIds = [random_img_id])
img = coco.loadImgs(imgId)[0]
print("Image details \n",img)

I = io.imread(img_dir + '/' + img['file_name'])
plt.axis('off')
plt.imshow(I)
plt.show()

#get damage annotations
annIds = coco.getAnnIds(imgIds=imgId,iscrowd=None)
anns = coco.loadAnns(annIds)

#Plot damages
plt.imshow(I)
plt.axis('on')
coco.showAnns(anns, draw_bbox=True )
plt.show()

#get parts annotations
mul_annIds = mul_coco.getAnnIds(imgIds=imgId,iscrowd=None)
mul_anns = mul_coco.loadAnns(mul_annIds)

# Create a dictionary between category_id and category name
category_map = dict()

for ele in list(mul_coco.cats.values()):
    category_map.update({ele['id']:ele['name']})

#Create a list of parts in the image
parts = []
for region in mul_anns:
    parts.append(category_map[region['category_id']])

print("Parts are:", parts) 

#Plot Parts
I = io.imread(img_dir + '/' + img['file_name'])
plt.imshow(I)
plt.axis('on')
mul_coco.showAnns(mul_anns, draw_bbox=True )
plt.show()