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

How to duplicate a class to train a subset class #763

Open
ClimbsRocks opened this issue May 8, 2019 · 0 comments

Comments

Projects
None yet
1 participant
@ClimbsRocks
Copy link
Contributor

commented May 8, 2019

Questions and Help

I'm able to train on my own custom dataset successfully (thanks to everyone who's posted helpful comments and code snippets on GitHub Issues these past months!).

I've seen repeated comments about selectively performing model surgery to take advantage of already-trained classes. For instance, let's say I'm training on two classes: person and dancing_person. It seems likely that dancing_person can learn a lot from the already-trained person class, since it's a subset of the person class.

Goal: Create a new class that uses the weights from an existing class, while still keeping the original class. If it matters, the new class is a subset of the existing class, but I want to get predictions from both.

Question: Can anyone point me to example code where this is done? Or even just point me to the relevant parts of the codebase with some ideas of what to do?

Thanks in advance if you can help!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.