tuyki / tt_rnn Goto Github PK

Hi

I downloaded the dataset and set up the paths.
However, I could not guess the format of a split.txt file. Could you give an example of what would be a split.txt file?
split_read = splits_path + classes[k] + '_split.txt'

Very interesting work! I have been trying to re-run UCF11 experiment after reading your paper. I have tried a number of formats to prepare the data to fit load_data API. I believe I am missing some steps. What's the input shape for load_data? (nb_frames, 240, 320, 3) or (nb_frames, 120, 160, 3). I noticed that you mentioned to scale it to 160:120 using ffmpeg. Although the comment in the load_data API says the original 320:240. Perhaps if you would like to share a script for the data preparation that would be great. Alternatively, if you can point to your pickle files location, that would help too. Any help would be highly appreciated. Cheers!

Hi, i want to know how to process the data set when you do experiment. such as experiment on UCF11 data set, giving the mpg video to the variable "data_path" directly? but i used the video directly, the code was crashed at this_clip = pickle.load(read_in)[0], and I have processed the data set to frame, and resized the image to 160×120×3, it also crashed at there, although generate a sequence of RGB frames of size 160 × 120 from each clip at an fps(frame per second) of 24 in paper, behind the code of crashed, there are some comment says "of shape (nb_frames, 240, 320, 3)", how do i understand it?
as it well, you have processed the mpg video to a tensor like (nb_frames, 240, 320, 3), and use the function pickle.load() to load the processed data?

Very interesting work! I have been trying to re-run UCF11 experiment after reading your paper. I have tried a number of formats to prepare the data to fit load_data API. I believe I am missing some steps. What's the input shape for load_data? (nb_frames, 240, 320, 3) or (nb_frames, 120, 160, 3). I noticed that you mentioned to scale it to 160:120 using ffmpeg. Although the comment in the load_data API says the original 320:240. Perhaps if you would like to share a script for the data preparation that would be great. Alternatively, if you can point to your pickle files location, that would help too. Any help would be highly appreciated. Cheers!

Great work!
Maybe thoughts on updating this for the latest Keras?

Martin

In your paper, you said that you sample 6 random frames in ascending order of each clip as training data. However, in your code, you set GLOBAL_MAX_LEN = 1492, which means that you put the whole clip to the model as input data. I think this will cause different results. I feel uncertain about the presetting. I am looking forward to your reply. Thank you!