In this PR I've created a new notebook to analyse GLIP results based on its mAP. We discovered (not surprised) that GLIP did very badly when the class names are either non-descriptive (e.g. '0', '1' etc) or where they are very domain specific (e.g. a name of a bacteria)
Description
In this PR I've created a new notebook to analyse GLIP results based on its
mAP. We discovered (not surprised) that GLIP did very badly when the class names are either non-descriptive (e.g.'0', '1'etc) or where they are very domain specific (e.g. a name of a bacteria)