Mike Azatov

Thanks @wvangansbeke! In your paper you mention the following: ``` Table 3 reports the results when we overestimate the number of ground-truth classes by a factor of 2, e.g. we...

Hi @wvangansbeke , I've been experimenting with `self-label` part of the code and I ran into an interesting scenario. As `self-label` learns it converges from 10 classes to just 3,...

Or maybe freeze layers of the backbone at least somehow?

@littletomatodonkey , thanks for the tip. In my case I'm thinking to do transfer learning on PP-OCRV3. So I'm thinking to freeze the backbone of PP-OCRv3, MobileNetV1Enhance. Would something like...

@Yosiiiiiiiiiiiiiiii , did you figure out why you had this error?

Thanks, @WenmuZhou , is there any documentation about it? I'd like to learn more about it.

Hello @klshrinidhi, I have the exact same issue. Did you ever figure out what is the problem?

> > Hello @klshrinidhi, I have the exact same issue. Did you ever figure out what is the problem? > > I did not unfortunately. Please let me know if...

Did you find a workaround for this issue? I'm having the same problem

Thanks @anatolix , could you elaborate a bit on cells? And how you imagine changing the size of the cells? As I understand the [model](https://github.com/michalfaber/keras_Realtime_Multi-Person_Pose_Estimation/blob/master/model/cmu_model.py), the dimension reduction only happens...