My team has recently been trying to get Axon working to detect the FRC Cargo Balls. We have a dataset that we created on Supervise.ly but whenever we try training a model with it Axon always gets stuck on the same step. Have any other teams had similar issues / any ideas on how to fix it?
The training can be very slow and can seem unresponsive. How long did you leave it running? How many images are in your data set?
I found it helped to close other applications on my machine to make sure Docker had plenty of memory to run.
We got a dataset of 600 labelled images to run in 45 minutes which gave model that could identify balls. Definitely room for improvement - we are working on putting together a larger dataset now.
Maybe try a different computer. We did have trouble getting axon to run on a computer with only 4 gb of ram, switched to a computer with 8gb and it worked great.
Here’s the computer specs on what we are using to run the model.
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.