yrodriguezmd / auto_annotate

Automate approach to label images for object detection using TensorFlow

Auto-annotate

Welcome to the auto-annotate images for TensorFlow object detection!

You are tired to label your images by hand to work with object detection? So, this project will make your life easier, just annotate some images and let the machine do the rest for you!

Requirements

• You will need to clone the TensorFlow repository
• Install the dependencies for object detection

note: This project is compatible with TF>=1.4

How to run

• Copy and paste the file generate_xml.py and visualization_utils.py into the research/object_detection/utils in the tensorflow repo.
• Change the xml path in generate_xml.py to put your own local path.
• Add the images you want to label into the images folder

TF < 2

• Add your pre-treined model (as a fronzen inference graph) and label map into the 'graphs' folder.
• Inside the auto_annotate folder run: python3 scripts/detection_images.py

TF > 2

• If you have TF 2.x, just run the following command:

python3 scripts/detection_img_tf2.py --saved_model /path-saved-model --labelmap /path-label-map.pbtxt --imgs /path-of-the-imgs

• If it runs correctly, you will see the inference results and the xml in your respective folders!

Any trouble?

If you have trouble or doubt check my tutorial on medium. You can also open an issue and I'll hep you!