Fixing Object Detection Models with Better Data

Object detection tasks can be particularly tedious to debug. If you've worked with large object detection datasets in the past, chances are you've run into incorrectly labelled data or data that's missing labels that end up killing your evaluation metrics. Identifying these issues usually involves manually inspecting the individual problematic examples in your dataset. The other issue with object detection is that these models usually output multiple detection boxes for a given image, and evaluation metrics have to be calculated based on different threshold parameters that control the strictness of our detection criteria. The most important parameter in this case, is the Intersection over Union (IoU) threshold.