Customizations of tensorflow-based applications.

Currently available:

• efficientdet - Docker image for EfficientDet: Scalable and Efficient Object Detection
• image-segmentation-keras - Docker image for image segmentation using Keras
• image classification - Docker images for training image classification models and making predictions with them (uses TensorFlow 1)
• image classification2 - Docker images for training image classification models and making predictions with them (uses TensorFlow 2)
• object detection - for building Docker images of TensorFlow's object detection framework
• tflite_model_maker - library for training models and making predictions using tflite's model maker
• tfutils - library for some common TensorFlow operations

For generating tfrecords, please see the wai.annotations library, which can convert to and fro various file formats.