I am writing .tfrecord-files to train an object detection model using TensorFlow in google colab. The data I use comes in the coco-json format.
There are multiple objects (labels) in each image from the coco-dataset.
My question is the following: does every single label have to be written as a seperate tf.train.Example with image info etc., or is there a way to create a list of features to one image info, so that there is only one tf.train.Example per image?