How can I convert my custom dataset to the prefetch dataset type and add features in the dataset

Thanks for picking up my thread,
I have a major doubt which is the basement for the assignment.
In the course we used the calltech bird dataset
But I didn’t really understand what each term meant, Like for example in the horses or humans assignment, we had the prefetch dataset as like this

<PrefetchDataset shapes: ((224, 224, 3),  ()), types: (tf.float32, tf.int32)>
                                               |__________|   |_|
                                                        |                 |
                                               Image tensors    Labels

Just like this can anyone explain me what does these terms mean

<PrefetchDataset shapes: {bbox: (4,), image: (None, None, 3), image/filename: (), label: (), label_name: (), segmentation_mask: (None, None, 1)}, types: {bbox: tf.float32, image: tf.uint8, image/filename: tf.string, label: tf.int64, label_name: tf.string, segmentation_mask: tf.uint8}>

Please I help me,


bbox: bounding box coordinates, four for each bbox, having float32 data type
image: bird image size (RGB images with different widths and heights, hence represented by None in the dataset), having uint8 data type
image/filename: name of the image file having string datatype
label: image label having int64 data type
label_name: name of the image label having string data type
segmentation_mask: the shape of the segmentation mask having uint8 data type

Thanks alot @AnkitSaini may god bless you… Is it necessary to do this calltech part because the zombie detector also works the same but this dataset has the bbox co-ordinates before itself.

You need bounding box coordinates to train the object detection model and hence it is necessary to have bounding box coordinates.