Dataset file size on disk is very large with 3D Array

Hi,

I have created my own dataset using the provided dataset loading script. It is an image dataset where images are stored as 3D Array with dtype=uint8.

The actual size on disk is surprisingly large. It takes 520 MB. Here is some info from `dataset_info.json`.

`{
"description": "",
"citation": "",
"homepage": "",
"license": "",
"features": {
"image": {
"shape": [224, 224, 3],
"dtype": "uint8",
"id": null,
"_type": "Array3D",
}
},
"post_processed": null,
"supervised_keys": null,
"builder_name": "shot_type_image_dataset",
"config_name": "default",
"version": {
"version_str": "0.0.0",
"description": null,
"major": 0,
"minor": 0,
"patch": 0,
},
"splits": {
"train": {
"name": "train",
"num_bytes": 520803408,
"num_examples": 1479,
"dataset_name": "shot_type_image_dataset",
}
},
"download_checksums": {
"": {
"num_bytes": 16940447118,
"checksum": "5854035705efe08b0ed8f3cf3da7b4d29cba9055c2d2d702c79785350d72ee03",
}
},
"download_size": 16940447118,
"post_processing_size": null,
"dataset_size": 520803408,
"size_in_bytes": 17461250526,
}`

I have created the same dataset with tensorflow_dataset and it takes only 125MB on disk.

I am wondering, is it normal behavior ? I understand `Datasets` uses Arrow for serialization wheres tf uses TF Records.

This might be a problem for large dataset.

Thanks for your help.