Scientific Contribution on Smart Data Management in an Iterative Context

This scientific report describes a thourough experimental study and contributions on active learning (AL) and class incremental learning (IL) for 2D object detection.
This work aims to provide more insights on the dataset construction phases, both for AL and IL scenarios.
We systematically test methods on several datasets to ensure general and reliable results. Moreover, we target the driving use-case which is more challenging than academic picture dataset in common use.