Segment Benchmark Data

[arinkulshi-skylight]

medical prescription and medical report benchmark data needs to be segmented.

• Look at all individual image files in medical prescription and medical report.
• segment all files in each dataset accordingly
• Ensure the format of ground truth is key: value and the labels is formatted as below
• Add any relevant code/changes to format the new files.

Ground Truth Format:
"box_b_employer_identification_number": "73-3075856",

Labels File Format

{
   "label": "box_b_employer_identification_number",
   "type": "text",
   "color": "250,147,170"
 },

[arinkulshi-skylight]

https://github.com/CDCgov/ReportVision/pull/318/files#diff-9bfc69e92588607e01d62a32ccb7663f224b2b985bc3f0c1b239e7a1008343ae

I have created a script to format the medical reports and medical prescription files here. This file will convert data from FHIR --> Key:Value as seen below.

[arinkulshi-skylight]

https://drive.google.com/drive/folders/1WS2FYn0BTxWv0juh7lblzdMaFlI7zbDd?usp=drive_link all segmented data + labels as well as metrics can be found here.