Skip to content

Conversation

@Poker-sang
Copy link
Contributor

For greater efficiency and better cross-platform.

Poker-sang added a commit to Poker-sang/Kagami that referenced this pull request Nov 16, 2022
@d-franklin
Copy link

There are still issues with the code. I've fixed an issue with the regions being the incorrect size and location and I'm working on fixing another issue related to actual detection.

Just a heads up for anyone looking to run yolov5-net on a non-windows platform and is looking at this pull for a possible solution.

@Poker-sang
Copy link
Contributor Author

Poker-sang commented Nov 22, 2022

There are still issues with the code. I've fixed an issue with the regions being the incorrect size and location and I'm working on fixing another issue related to actual detection.

Just a heads up for anyone looking to run yolov5-net on a non-windows platform and is looking at this pull for a possible solution.

I am sorry that there's still a problem. It seems like the predict rectangle, is based on the resized input image, but not based on the original one. So I think the simplest fix for this is to resize the predict rectangle again.

@Poker-sang
Copy link
Contributor Author

Poker-sang commented Dec 12, 2022

Now I fix the bug that made predict rectangle distorted, through removing the code below:

var gain = Math.Min(xGain, yGain); // gain = resized / original

Now I thought it can behave just like the main repository.

@kkarahainko kkarahainko merged commit dac6e69 into techwingslab:master Mar 23, 2023
@d-franklin
Copy link

There is still bugs in this code, I ran out of time so I couldn't fix them and I haven't touched it in 6 months so I can't remember what unfortunately, but they are there.

@kkarahainko
Copy link
Contributor

bugs are always there ) no worries, thanks for contribution!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

3 participants