EOGrady21 / vprr

Video Plankton Recorder Data Processing
https://eogrady21.github.io/vprr/
Other
2 stars 1 forks source link

additional checks #53

Closed kevinsorochan closed 11 months ago

kevinsorochan commented 11 months ago

**Is your feature request related to a problem?

In the development phase, there have been some problems with duplication and missing image paths in the final aid files.

Describe the solution you'd like There should be a check that scans for this in the final version. This could even be directly incorporated into vpr_autoid_create()

EOGrady21 commented 11 months ago

I updated vpr_autoid_check() based on our most recent checks

It makes no edits or changes to the aid files but outputs a text file with the results of 3 checks. (1. Checks for empty files, 2. Check for duplicated ROIs by hour, 3. checks for missing ROIs by hour between automated results and new_aids output after manual reclassifications)

It can be run like this

vpr_summary <- read.csv('c:/data/CAR2022299/vpr_metadata_CAR2022299.csv')
# day_hour column should look like 'd123.h12'

vpr_autoid_check(new_autoid = './new_autoid/',
                 original_autoid = 'c:/data/CAR2022299/VPR/autoid_cnn_scr/',
                 cruise = 'CAR22022299',
                 dayhours = unique(vpr_summary$day_hour))

and output looks like this CAR22022299_aid_file_check.txt