Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Extract Trainset Responding Slow #34

Open
lizimai opened this issue Dec 10, 2021 · 3 comments
Open

Extract Trainset Responding Slow #34

lizimai opened this issue Dec 10, 2021 · 3 comments

Comments

@lizimai
Copy link

lizimai commented Dec 10, 2021

Hi! I encounter the slow response issue when I am using the graphic tool of anTrax: Extract Trainset.

  1. When I try to change between images in antrax, the software start to respond really slow (> 30s) after the 50th image.
  2. When I try to switch between movies via the "Tracklets" option in the top bar menu, the process of "Preparaing validation datset, loading images..." takes around 5-10 mins.

The first issue was also observed in a different higher memory machine. However, I don't know if the second issue is due to the low memory on my machine.

Thanks a lot!


Information of my machine:
MacBook Pro 2017,
Processor 2.3 GHZ Dual Core Intel Core i5
Memory 8 GB 2133 MHz LPDDR3
MacOS Catalina 10.15.7
Matlab 2020a
Python 3.7.10

@asafgal
Copy link
Contributor

asafgal commented Dec 10, 2021

Hey Zimai,

By default, the software preloads the images from 50 tracklets when the GUI opens. After that, it will load the images only when you ask for a new tracklet, hence the delay (although 30sec is too much I must say). The cli does not allow you to change this setting, but you can launch the gui from the matlab prompt with:

validate_classifications(<path-to-expdir>, 'N', <num-of-tracklets-to-preload>);

Obviously, the price you pay is a delay in opening the GUI. You can also choose a specific video range to draw tracklets from, or that are classifies with a specific ID (if you are doing a second round training):

validate_classifications(<path-to-expdir>, 'N', <num-of-tracklets-to-preload>, 'movlist', <list-of-video-inices>, 'id', <ID>);

This is not related to your machine, but to inefficiencies in the way anTraX (via MATLAB) stores and access tracklet data.

Let me know if it helps!

@lizimai
Copy link
Author

lizimai commented Dec 16, 2021

Thanks a lot Asaf! However, I think the bigger problem is the point 2 that I listed: the loading process when switching between videos in extract-tracklet takes >10 mins and thus it takes very long for me to build a transet. I am not sure if this can be improved by the machine?

@asafgal
Copy link
Contributor

asafgal commented Dec 16, 2021

Can you give me some info about your experiment / data? How many videos you have, how many colonies/ants in the frame. how long is each video, whats the order of magnitude of the number of tracklets per video? The delay in loading a file can be a result of a very fragmented dataset (a file containing very large number of short tracklets).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants