For the Training Image Labeler, can you include images without the structure-of-interest to improve the RCNN training?

For the Training Image Labeler, can you include images without the structure-of-interest to improve the RCNN training?

Answer 1

[Nick Choi]

You could provide more images without the structure of interest to the Training Image Labeler but there is no guarantee that it would improve the performance of the RCNN. It depends on the data set but including more images without the structure of interest can have trade offs as mentioned in this doc link: https://www.mathworks.com/help/vision/ug/train-a-cascade-object-detector.html#btu4ez9-1