Countix-AV is a dataset for repetitive action counting by sight and sound created by repurposing the Countix dataset.
It is created by selecting 19 categories from Countix for which the repetitive action has a clear sound, such as clapping, playing tennis, etc. The dataset contains 1,863 videos, with 987, 311 and 565 for training, validation and testing.
The authors maintained the original count annotations from Countix and kept the same split (i.e. training, validation, or testing) for each video.
Paper | Code | Results | Date | Stars |
---|