Adding Support for Loading Pascal VOC dataset format to tf.data about chitra HOT 4 CLOSED

I have used this feature to convert to tf.Record file using the API provided by tensorflow itself. In that one needs to run a python script from command line with apt arguments to get tf.Record file.
Now, can you clarify some of my questions:

1. In tf.data you want to create same Record file or or something else ?
2. Are you trying to implement the same thing implemented in that script, in that function. Or one can find a way to use that same script file provided by tensorflow, in backend implementation of this function ?

I want to build a function that accepts two arguments - Pascal VOC dataset path and then loads it in tf.data format directly (without TFRecord dataset):

def load_dataset(image_path, xml_path):
     # preprocess
     # convert data to tf.data
     return image_filename, image, bounding_boxes, labels

You can read more about tf.data here

Do let me know if you need any help.

from chitra.

Thanks, first would like to do a bit research on this.

from chitra.

I have used this feature to convert to tf.Record file using the API provided by tensorflow itself. In that one needs to run a python script from command line with apt arguments to get tf.Record file.
Now, can you clarify some of my questions:

1. In tf.data you want to create same Record file or or something else ?
2. Are you trying to implement the same thing implemented in that script, in that function. Or one can find a way to use that same script file provided by tensorflow, in backend implementation of this function ?

from chitra.

Stale issue message

from chitra.