assess_overlap: assess_overlap
In jthinke/vvipr: Verify VIAME Predictions
View source: R/assess_overlap.R
assess_overlap R Documentation
assess_overlap
Description
The workhorse function in the vvipr app. Users should not need to call this directly. The function is based on sf geometries for assessing the performance of image classification models implemented with VIAME software (https://www.viametoolkit.org/)
Usage
assess_overlap(truth, prediction, conf.thresh = 0.2, over1 = 0.5, over2 = 0.5)
Arguments
truth
A path identifying the user-selected .csv data file representing the truth annotations as output by VIAME software for the set of images under analysis
prediction
A path identifying the user-selected .csv data file representing the model predictions as output by VIAME software for the set of images under analysis
conf.thresh
A numeric value from 0 to 0.99 that specifies the minimum value for including model predictions in the analysis. Predictions with confidence
levels below the threshold will be eliminated from the analysis.
over1
A numeric value from 0 to 1 that specifies the minimum proportion of a truth annotation that must be covered by a model prediction. Predictions
that do not meet this threshold will be assigned a status of 'false positive'
over2
A numeric value from 0 to 1 that specifies the minimum proportion of a prediction area that must overlap with a truth annotation. Predictions that
do not meet this threshold will be assigned a status of 'false positive'. Note that predictions that fail to meet the 'over1' threshold, but do meet the 'over2
threshold will be assigned a 'true positive' status
Details
The function requires input of two files, as output by VIAME, and the setting of 3 thresholds for assigning a model prediction as a false positve.
The images used for the truth and prediction files should be the exact same.
Value
The function returns a list containing a data frame of model results, a list of class sf with geometries for each image and class, a vector of
polygon IDs for all false positives that were detected, and a data frame with CLASS and INDEX headers identifying classes in the data.
jthinke/vvipr documentation built on Nov. 3, 2023, 8:39 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
View source: R/assess_overlap.R
| assess_overlap | R Documentation |
assess_overlap
Description
The workhorse function in the vvipr app. Users should not need to call this directly. The function is based on sf geometries for assessing the performance of image classification models implemented with VIAME software (https://www.viametoolkit.org/)
Usage
assess_overlap(truth, prediction, conf.thresh = 0.2, over1 = 0.5, over2 = 0.5)
Arguments
truth |
A path identifying the user-selected .csv data file representing the truth annotations as output by VIAME software for the set of images under analysis |
prediction |
A path identifying the user-selected .csv data file representing the model predictions as output by VIAME software for the set of images under analysis |
conf.thresh |
A numeric value from 0 to 0.99 that specifies the minimum value for including model predictions in the analysis. Predictions with confidence levels below the threshold will be eliminated from the analysis. |
over1 |
A numeric value from 0 to 1 that specifies the minimum proportion of a truth annotation that must be covered by a model prediction. Predictions that do not meet this threshold will be assigned a status of 'false positive' |
over2 |
A numeric value from 0 to 1 that specifies the minimum proportion of a prediction area that must overlap with a truth annotation. Predictions that do not meet this threshold will be assigned a status of 'false positive'. Note that predictions that fail to meet the 'over1' threshold, but do meet the 'over2 threshold will be assigned a 'true positive' status |
Details
The function requires input of two files, as output by VIAME, and the setting of 3 thresholds for assigning a model prediction as a false positve. The images used for the truth and prediction files should be the exact same.
Value
The function returns a list containing a data frame of model results, a list of class sf with geometries for each image and class, a vector of polygon IDs for all false positives that were detected, and a data frame with CLASS and INDEX headers identifying classes in the data.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.