The Google Cloud Storage URI of the input csv file.
The URI must start with
The format of the input csv file should be one image per
line. In each line, there are 6 columns.
1. image_uri 2, image_id
3. product_id 5, product_category 6, product_display_name 7,
Columns 1, 3, 4, and 5 are required, other columns are
optional. A new ProductSet/Product with the same id will be
created on the fly if the ProductSet/Product specified by
product_set_id/product_id does not exist.
The image_id field is optional but has to be unique if
provided. If it is empty, we will automatically assign an
unique id to the image.
The product_display_name field is optional. If it is empty,
a space (" ") is used as the place holder for the product
display_name, which can be updated later through the
If the Product with product_id already exists, the fields
product_display_name, product_category and labels are
If a Product doesn't exist and needs to be created on the
fly, the product_display_name field refers to
the product_category field refers to
and the labels field refers to [Product.labels].
Labels (optional) should be a line containing a list of
comma-separated key-value pairs, with the format
The bounding_poly (optional) field is used to identify one
region of interest from the image in the same manner as
CreateReferenceImage. If no bounding_poly is specified, the
system will try to detect regions of interest automatically.
Note that the pipeline will resize the image if the image
resolution is too large to process (above 20MP).
Also note that at most one bounding_poly is allowed per
line. If the image contains multiple regions of interest,
the csv should contain one line per region of interest.
The bounding_poly column should contain an even number of
comma-separated numbers, with the format
"p1_x,p1_y,p2_x,p2_y,...,pn_x,pn_y". Nonnegative integers
should be used for absolute bounding polygons, and float
values in [0, 1] should be used for normalized bounding