Google Code-in

Food101 dataset of TensorFlow Datasets is having only the train split of the dataset. However the original source of the dataset contains 25250 validation data which are being completely ignored in TFDS.

The original source can be found here. https://www.vision.ee.ethz.ch/datasets_extra/food-101/

The task is to rewrite the Dataset Handler so that it parses the validation data and returns a SplitGenerator on the validation data along with the training data.

Important Links:

• https://www.tensorflow.org/datasets/add_dataset
• https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/image/food101.py
• https://github.com/tensorflow/datasets/blob/master/tensorflow_datasets/image/food101_test.py