Error running code about v2e HOT 13 CLOSED

from v2e.

It's possible there is a bug with the options since it was hard to test all combos. What was your command line exactly? Did you supply or the other of ERROR:main:specify either --overwrite or --unique_output_folde Thanks

Thank you for your reply.
I used what you provided sample input data (included SuperSloMo model and tennis.mov).
I'm going to put these in a folder called Input.
Then, Use the command in cmd "python v2e.py -i input/tennis.mov --timestamp_resolution=.005 --dvs_exposure duration 0.005 --output_folder=output/tennis --overwrite --pos_thres=.15 --neg_thres=.15 --sigma_thres=0.03 --dvs_aedat2 tennis.aedat --output_width=346 --output_height=260 --stop_time=3"
I got the error.

from v2e.

Your command looks fine to me. I see you are using Windows to run v2e, could it a path problem? Maybe you can try to substitute input and output folders with absolute paths and try again?

from v2e.

Your command looks fine to me. I see you are using Windows to run v2e, could it a path problem? Maybe you can try to substitute input and output folders with absolute paths and try again?

I solved the problem!
Remove "--overwrite" from the command line.Thanks for the tip.
One more question，I found that the generated video is not the original video before the conversion(In terms of time and speed of the video).If I want to get the same length and frame rate of video, how do I set the parameters?

from v2e.

from v2e.

Which generated video? Do you mean the DVS video? If you refer to DVS video, yes, it will not be same length as input video in general since it might be exposed by constant_duration, constant_count or area_count methods. And frame rate of constant_duration DVS frame output video is (I recall) taken from frame rate of input video, but the DVS video is slowed down by the frame interpolation. Could you provide a bit more specific example? And look carefully at the extended usage output from --help invocation. Thanks

Yes，it is dvs-video.avi.
I used the example video you mentioned(tennis.mov ),it has 25 seconds.However, The generated DVS video is only 19 seconds long and the frame rate is noticeably slower.
Well,let me take a closer look at the help.

from v2e.

from v2e.

It could well be a bug, thanks for noticing. I'm extremely busy now with coding up something for the Telluride workshop and can't spend time right now to look. I can look after next week. Let me know if you need some hints.

Hi, thank you for your clarification.
Now I have another problem.I have some data in .aedat2.0 format and want to convert it to .txt format.
Is there a script for such a simple format conversion?(Actually, I tried and failed.My answer seems to be (x,y), no t)

from v2e.

@xlj199

A possible way to do this is through the DV python bindings, it implements AEDAT2.0 API. See here: https://gitlab.com/inivation/dv/dv-python/blob/master/README.md

It doesn't support the reading for the frames, but I guess that's not a problem here.

For a relatively naive yet working solution for a standard AEDAT 2.0 file, you can decode the file using this function:
https://github.com/SensorsINI/jaer-control/blob/master/jaercon/procaedat.py#L192

The bad side of this solution is that it loads the entire recording into the memory.

But you need to write the code to drop all the data into a text file.

from v2e.

@duguyue100 I implemented the idea, but the parameters T seemed hard to understand(It's not starting from zero, and it has no units).

from v2e.

The unit for time is microsecond and in AEDAT2, it's coded into a INT64 number.
As the T not starting from 0, I think it's normal because the events are uniformly distributed between 0 and 2nd frame's time.
At least it should be a number that is close to 0.
Make sure you decode the AEDAT format correctly as in the reference code above.

from v2e.

from v2e.

I hope you find the solution, close the issue for now.

from v2e.