Expected 11 columns, but found 3 columns only

[Joel Anderson]

Hi,

I am trying to import a csv file into the vision AutoML.   It keeps complaining that it needs 11 columns.  I've also tried some other variants, all require 11 columns.  I do want to use the bounding boxes.  Has the format changed?

AutoML doc says the format is :

[set,]image_path[,label,x1,y1,,,x2,y2]
TRAIN,gs://My_Bucket/sample1.jpg,cat,0.125,0.25,,,0.375,0.5
VALIDATE,gs://My_Bucket/sample1.jpg,cat,0.4,0.3,,,0.55,0.55
TEST,gs://My_Bucket/sample1.jpg,dog,0.5,0.675,,,0.75,0.875

Here's some example lines it complains about:

TRAIN,gs://mytensorflowdata/CanVideo 50.jpg,sodacan,0.378125,0.10138888888888889,,,0.61796875,0.8708333333333333

TRAIN,gs://mytensorflowdata/CanVideo 50.jpg,sodacan

gs://mytensorflowdata/CanVideo 50.jpg

Am I doing something wrong?  What are the 11 columns?

Cheers,

Joel

[Ali (Cloud Platform Support)]

Hi,

Following the documentation to prepare your data in CSV format, each row should have the following format:

First column has to hold the value TRAIN, VALIDATE, TEST or UNASSIGNED. 

The second column if the GCS path of your image.

The third column is the label of your image. 

The remaining columns hold the bounding box in either of the two formats:

-(x_relative_min, y_relative_min,,,x_relative_max,y_relative_max,,)

        -(x_relative_min,y_relative_min,x_relative_max,y_relative_min,x_relative_max,y_relative_max,x_relative_min,y_relative_max)

As displayed in the first format of the bounding box, even if a column is empty, you still have to include a comma with nothing indicated after it. In other words, each row is required to have 11 columns even if some columns for some rows are empty for the upload to properly work. 

[Joel Anderson]

Thank you!

This worked.