Assigned
Status Update
No update yet.
Comments
gs...@google.com
Hello Alexander,
You mention other images of tables; what other images have you tried? Have you tried cases where the table did not fill the entire page, but less than a third? BlockType TABLE is defined as "Table block." on the "Method: images.annotate" documentation page. [1]
What extra information would this BlockType value as TABLE would bring for you? In fact, you are already aware that what you submit to the API call is the image of a table.
A few images that you tried would be of help in reproducing this issue on our side. The image you attached is particular in that the table fills the entire page.
[1]https://cloud.google.com/vision/docs/reference/rest/v1/images/annotate#blocktype
You mention other images of tables; what other images have you tried? Have you tried cases where the table did not fill the entire page, but less than a third? BlockType TABLE is defined as "Table block." on the "Method: images.annotate" documentation page. [1]
What extra information would this BlockType value as TABLE would bring for you? In fact, you are already aware that what you submit to the API call is the image of a table.
A few images that you tried would be of help in reproducing this issue on our side. The image you attached is particular in that the table fills the entire page.
[1]
pu...@gmail.com
Hi,
I cropped the table from full invoice in order to have small result data set for testing porposes, usually the table comes integrated in bigger document.
I need a block_type value because:
1.I want to get table and other data from full invoice and parse them separately with different logic.See bill.jpg
2.I have invoices with multiple tables on them and additional data so i need to distinguish the tables(or table like structures) and parse them accordingly. See super.jpg
I am attaching another image, in all those 3 images i get only block_type = 1 for all blocks.
I am testing with code that taken fromhttps://cloud.google.com/vision/docs/fulltext-annotations
just adding the print under the block 'for'.
```
for block in page.blocks:
print('BLOCK TYPE: ' + str(block.block_type))
```
I cropped the table from full invoice in order to have small result data set for testing porposes, usually the table comes integrated in bigger document.
I need a block_type value because:
1.I want to get table and other data from full invoice and parse them separately with different logic.See bill.jpg
2.I have invoices with multiple tables on them and additional data so i need to distinguish the tables(or table like structures) and parse them accordingly. See super.jpg
I am attaching another image, in all those 3 images i get only block_type = 1 for all blocks.
I am testing with code that taken from
just adding the print under the block 'for'.
```
for block in page.blocks:
print('BLOCK TYPE: ' + str(block.block_type))
```
gs...@google.com
Your specific issue has been made known to Engineering, who will address it in due course. No estimated time to resolution has been set. Meanwhile, you may follow developments in this thread.
gs...@google.com
OCR does not provide "table" block type. A table parsing enhancement to DOCUMENT_TEXT_DETECTION might be implemented in future. As yet, there is no established release date. Progress with get recorded in this thread.
jm...@google.com
ja...@diverseprogrammers.com
[Deleted User] <[Deleted User]>
Suprisingly, the REST API doesn't provide structure information or blocktype while rpc does, I wonder why is that the case.
js...@gmail.com
Any update on this?
se...@doncelsoft.com
I'm also interested on this
sr...@gmail.com
I am also interested in this, AWS Textract is giving option to find the table in the images. Can we expect the same here also
hr...@gmail.com
I am also interested in this, I am trying to compare this with Azure and AWS.
sw...@gmail.com
Hello Google, Why is there no update on this issue? Please update!
ni...@aramix.ai
I am interested too
fr...@aramix.ai
I am interested too
Description
Problem you have encountered:
BlockType 'TABLE' is never returned from REST API I receive only TEXT
Attached one of the images i tried to get table from.
What you expected to happen:
To receive a BlockType TABLE
Steps to reproduce:
I run the default example on pyton:
```
from google.cloud import vision
client = vision.ImageAnnotatorClient()
with io.open(path, 'rb') as image_file:
content = image_file.read()
image = vision.types.Image(content=content)
response = client.document_text_detection(image=image)
```
Other information (workarounds you have tried, documentation consulted, etc):