Skip to content

Instantly share code, notes, and snippets.

@chricke

chricke/create_records.sh

Last active Mar 28, 2019
Embed
What would you like to do?
create TFRecords
lumi dataset transform --type csv --data-dir data/ --output-dir tfdata/ --split train --split val --only-classes=table
@jyoti59n

This comment has been minimized.

Copy link

@jyoti59n jyoti59n commented Dec 25, 2018

Hi,

I'm unable to run the above luminoth script - trying to replicate the run on the same set of train/val images that you have used. I get the following message:

'Columns missing from CSV: {}'.format(missing_keys)
luminoth.tools.dataset.readers.base_reader.InvalidDataDirectory: Columns missing from CSV: {'xmin', 'ymax', 'xmax', 'image_id', 'label', 'ymin'}

Also notice that this error throws up for a random order of the csv columns for different runs. e.g. in an earlier run the message was
luminoth.tools.dataset.readers.base_reader.InvalidDataDirectory: Columns missing from CSV: {'xmax', 'ymin', 'image_id', 'xmin', 'ymax', 'label'}

Would you be able to suggest what I'm doing wrong here? Thanks.

@goodhamgupta

This comment has been minimized.

Copy link

@goodhamgupta goodhamgupta commented Jan 16, 2019

Hi @jyoti59n,

There is a problem with the given train.csv and val.csv files. The header rows are missing because of which you get the missing columns error(Check the _check_csv function of the csv_reader in luminoth here).

Add the following to the first line of both the flies:

image_id,xmin,ymin,xmax,ymax,label

It should work after this change.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.