Sampling Images for layout annotation

[10zinten]

• Top 5 and middle/bottom 5 images from every imagegroup.
• top offset: 5

[kaldan007]

top offset can be look at the property volumePagesTbrcIntro in the database, for instance in http://purl.bdrc.io/resource/I0886.ttl for the bottom we don't have any other option then using list spliting. To get image list we can either look on S3 or through https://iiifpres.bdrc.io/il/v:bdr:{img_grp_id}

[kaldan007]

I think its better to go with 10 from top and 10 from bottom as we can ignore blank image using the ignore button.

[ta4tsering]

okay will look into it

[ta4tsering]

the link below is the script that takes a work_id, gives back a dict with imagegroup as key and s3_key of the top 10 and bottom 10 images from that imagegroup, excluding the tbrc intro images from the top. image sampling for layout analysis