Detailed changelog

[xternalz]

Hi, would you be able to provide more a detailed changelog about going from V2 to V3? More specifically, I would like to know whether the annotations existed in V2 were re-annotated in V3?

Thanks.

[nalldrin]

Hi, there isn't a current plan to provide a detailed changelog for V2->V3. The main change is simply a lot more box and image-level groundtruth. There is, however, a small fraction of image-level groundtruth that changed due to internal cleanups (of note we had some cases where the box drawing or verification task indicated that the image-level groundtruth was wrong, we removed the groundtruth for these cases because an audit revealed a mix of correct and incorrect cases; we plan to revisit these cases for V4 release). There is also a fraction of the groundtruth that were re-annotated (one example is from crowdsourcing).