About frames number

[ApolloRay]

In this paper, it mentions that "For very high-resolution images, we limit the maximum number of grids to 49". For LongVA, each frame of the video is treated as an image 336 * 336 * N. I'm not sure whether it means the maximum number of frame = 49 ?

[jzhang38]

The max number of N is 49. That is just a limit we use in the code. I do not really know if there is such high res image in the training set. Normally the res for OCR data set is around 1K and the rest is around 500P.