How to work with COCO object detection datasets in Tensorflow?

Issue

I’m super new to this topic of object detection and TensorFlow and I was wondering how can I load this file as a TensorFlow dataset? I’ve used pandas to read it as a DataFrame but I can’t parse it to a TensorFlow dataset. I’ve tried this

train_ds = tf.data.Dataset.from_tensor_slices(
  (
    tf.cast(dataframe['file'].values, tf.string),
    tf.cast(dataframe['width'].values, tf.int8),
    tf.cast(dataframe['height'].values, tf.int8)
  ),
  (
    # tf.cast(dataframe['annotations'].values, tf.string)
    # here I want to cast the annotations to the dataset but I don't 
    # know how
  )
)

but could not figure out how to get the annotations to work here… Any help with that?

Solution

Looks like your data is in JSON format,
Directly use tf.io.decode_json_example library to read the json values.

import tensorflow as tf
tf.io.decode_json_example([
    [example_json, example_json],
    [example_json, example_json]]).shape.as_list()

For more details on the library find here.

Answered By – TFer

This Answer collected from stackoverflow, is licensed under cc by-sa 2.5 , cc by-sa 3.0 and cc by-sa 4.0