about video classification

[Auroralyxa]

Hi, I have three operational questions about video classification 1.I don't know how to resize the video to the short edge size of 320,Could you please provide me with the processed video, my email is 2273740426@qq.com 2.Can I replace the video with time series image data? The time series image data refers to a large number of images in June and September every year from 2017 to 2021. If possible, which part should I modify? Thanks in advance

[Andy1621]

1. You can find Kinetics400 here.
2. I am not sure of the meaning of time series image data. Can you provide some examples? And what is the purpose to use them?

[Auroralyxa]

Thank you for your reply. I am studying crop classification. According to different phenological periods, the characteristics of crops are different, that is to say, the crop images obtained at different times are different, so it is necessary to add temporal features.
The size of each image is different. Example images are as follows:

If I want to replace the video with time-series images, which part should I modify?

[Andy1621]

In my opinion, your dataset seems like a frame-based video dataset. You can move all the images belonging to the same crop into one directory, and rename them as 00001.png/jpg, 00002.png/jpg... (the number increase with the time) Then you can prepare your meta file like this. For example, if you have 32 time-series images for each crop, and the label are 0 and 1,

# path,frame_number,label
crop1,32,0
crop2,32,0
crop3,32,1
crop4,32,1

[Auroralyxa]

Thank you, I will try as you say

[Andy1621]

As there is no more activity, I am closing the issue, don't hesitate to reopen it if necessary.