--
You received this message because you are subscribed to the Google Groups "ActivityNet" group.
To unsubscribe from this group and stop receiving emails from it, send an email to activity-net...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/activity-net/0056e595-d0bc-4294-b36e-553df294dc05%40googlegroups.com.
Hi Junting,You're right. The big difference is the addition of new clips in the training / testing set from the Kinetics dataset.Interestingly these new clips also come with ground truth labels from the Kinetics vocabulary. I'm excited to see if anyone can improve performance using co-training on the two tasks, or other creative uses of the two label sets. Can someone win both the AVA and Kinetics challenges with a single model?Best regards,David
On Fri, May 15, 2020 at 7:21 AM Junting Pan <junti...@gmail.com> wrote:
Hi David,--I wonder what is the difference of the crossover challenge and AVA challenge besides the testing data?Best,Junting
El lunes, 27 de abril de 2020, 5:21:38 (UTC+2), David Ross escribió:Hi AllThe AVA-Kinetics crossover dataset is available for download herehttps://research.google.com/ava/. The arxiv paper describing it is coming very soon.If you spot any issues or have any questions, please let me know.Best regards,David RossGoogle Research
You received this message because you are subscribed to the Google Groups "ActivityNet" group.
To unsubscribe from this group and stop receiving emails from it, send an email to activi...@googlegroups.com.