0
$\begingroup$

I'm currently working on an object counting/density estimation task using low frame rate video (~2 fps) in a traffic setting. I've explored a lot of literature on both spatial methods (i.e. using only individual images) and spatiotemporal methods (i.e. utilizing the sequential nature of video frames); many of these are based on CNN-LSTM or transformer models. It seems to me that generally, if video is available, using spatiotemporal models can be more powerful.

My question is this: If I'm using pretty low frame rate video and the vehicles are moving 10-30 mph, is there a significant benefit to using a spatiotemporal over just spatial features? I ask because it seems more complicated to to the former, and if it doesn't bring significant benefits given my frame rate, I'd rather stick with something a bit simpler. Does anyone have any experience implementing either of these approaches for low frame rate video?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

It depends. The more information you input for a sample allows the model to better understand it. However, due to differences on data quality, dataset size, model size, etc., no one can guarantee your model will perform better. If your data are all of high quality, large amount, and if your model is of a proper structure and size, you are likely to get a performance gain.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .