search

tryolabs / luminoth

Deep Learning toolkit for Computer Vision.
https://tryolabs.com
BSD 3-Clause "New" or "Revised" License
2.4k stars 400 forks source link

How to adapte a dataset which use Microsoft/VoTT to export #251

Closed gongeek closed 5 years ago

gongeek commented 5 years ago

When I use Microsoft/VoTT to create my own dataset.

First I try export as TF Record . And then use lumi dataset merge to get one file . It can't be use in luminoth . It throw a error when train tensorflow.python.framework.errors_impl.InvalidArgumentError: Name: , Context feature 'depth' is required but could not be found image

Then I try export as Tensorflow Pascal Voc format .And then use lumi dataset transform --type pascal ... to transform it to .tfrecords format . But it throw Error don't find train.txt

image

How to adapte a dataset which use Microsoft/VoTT to export? Or Which Visual Object Tagging Tool you suggest to use ?

sleipniry commented 5 years ago

I have the same problem.Have you solved it?

gongeek commented 5 years ago

@sleipniry You can parse the img.json which Microsoft/VoTT exported to train.csv format . the format is like this image

https://github.com/tryolabs/luminoth/blob/master/luminoth/tools/dataset/readers/object_detection/csv_reader.py