Skip to content
Luke Genung edited this page Aug 1, 2019 · 2 revisions

Welcome to the tf-image-pipeline wiki!

Create an image pipeline to train a TensorFlow image classifier using video files.

Steps:

  1. video_to_mp4.py - Convert video files to MP4
  2. mp4_to_jpeg.py - Convert MP4 files to JPEG images for training TF classifier
  3. image_sort_gui.py - GUI to quickly label the training images by copying them into separate folders (manual labelling task)
  4. remove_background.py - Crop out the background of training images to reduce undesired features in training images

Lastly, the TensorFlow image classifier can be trained using: python retrain.py --image_dir ~/flower_photos

More TF info here: https://www.tensorflow.org/hub/tutorials/image_retraining

And here: https://codelabs.developers.google.com/codelabs/tensorflow-for-poets/#0

Clone this wiki locally