gamcoh / object-detection-tools

Tools for object detection annotations in machine learning

🔧 Object-Detection-Tools

Tools for object detection annotations in machine learning:

• Converts PASCAL VOC annotations to the _annotations.json format supported by the Cloud Annotations tool.
• Upload images and annotations to Cloud Object Storage Bucket.
• Converts PASCAL VOC annotations to the COCO json annotation format.
• Remove empty annotation files
• Remove some annotations in all the files

Instructions

Remove empty annotation

Remove empty xml annotation file. If a file does not have any object tag:

python remove_empty_annotation.py

Remove some annotation

This script removes an annotation from all of your xml files. If you want to remove all xml objects from your annotations files that have the label 'car':

python remove_annotation.py --name car

xml → json (Cloud Annotations Tool)

Convert xml annotations to json:

python convert_xml2json.py

xml → json (COCO json annotation format)

Convert xml annotations to coco json annotation format

python pascal_voc_xml2coco_json.py

Upload images and annotations to Cloud Object Storage Bucket

Install the Cloud Object Storage SDK:

pip install botocore==1.12.26 ibm-cos-sdk==2.3.2 ibm-cos-sdk-core==2.3.2 ibm-cos-sdk-s3transfer==2.3.2

Add your credentials to upload2bucket.py:

credentials = {
  'bucket': 'YOUR_BUCKET_NAME',
  'iam_url': 'https://iam.ng.bluemix.net/oidc/token',
  'resource_instance_id': 'YOUR_INSTANCE_ID',
  'url': 'YOUR_REGION_ENDPOINT',
  'api_key': 'YOUR_API_KEY'
}

Upload the images and the annotations:

python upload2bucket.py

Acknowledgments

• Thanks to @bourdakos1 for the great help
• Thanks to @CivilNet for the xml to coco script