phananh1010/PanoSalNet

Questions about fps

Closed this issue · 7 comments

I found that the frame rate you set is 30, I want to know how to set it when the frame rate is 24?

I did not see the problem with fps, can you elaborate the question?

I did not see the problem with fps, can you elaborate the question?
I want to generate a saliency map frame by frame on my data set, but my video’s frame rate is 24. I noticed that your saliency map is generated according to the timestamp, and the frame rate is 30, is it just modified like this:
get_fixation.py: https://github.com/phananh1010/PanoSaliency/blob/d67f0abbf323b2c6f98dfbf3055968bc802ab026/get_fixation.py#L15
line 15:DELTA -> 1.0/24
line 43:for t in np.arange(0, vlength, DELTA):

I see, the code you are viewing generate ground truth saliency map from head movement log. Specify a timestamp, it will find the closest head movement data within a threshold. Modifying DELTA as you suggested is a correct way.

Thank you very much!

In addition, I am also very interested in the part of your head movement prediction. I want to add more data dimensions, but when training with pytorch, I don’t know how to input so many features into it. Could you share the training code to us?

I don't have immediate access to the training code for the head prediction model now, unfortunately. However, the idea is to flatten all features and concatenate them together.

Thank you for your guidance!