Merged the posts
Search found 10 matches
- Sun Mar 13, 2022 7:39 pm
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
- Mon Mar 07, 2022 8:53 pm
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
Re: [Guide] Introducing - Phaze-A
I have moved over to refactoring the code of models as to have more flexibility on where specific layers are called. It's sadly not often the case that observations from custom models can be translated to Phaze-A settings, but some improvements such as EfficientNet v2 definitely will find their way ...
- Fri Mar 04, 2022 11:59 am
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
Re: [Guide] Introducing - Phaze-A
Thanks for answering. I'll note observations down in the following and keep them updated from time to time, On the technical level it would appear that using Shared FC (Full/Half) and/or not splitting the decoder will result in leakage of face features. That's because the weights of these models are...
- Sun Feb 13, 2022 3:00 pm
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
Re: [Guide] Introducing - Phaze-A
Merged the posts
- Sat Feb 12, 2022 3:28 am
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
Re: [Guide] Introducing - Phaze-A
Being able to load ImageNet weights to kickstart models is great. Thanks for implementing the feature and the entire Phaze-A trainer. Seems like a major improvement that I hope more people will play around with. Could you please give me some pointers on when the rest of the model has caught up to th...
- Thu Feb 03, 2022 3:06 pm
- Forum: Extract Discussion
- Topic: Multiple faces
- Replies: 1
- Views: 1716
Re: Multiple faces
You could sort the alignments by face. Ideally that would group faces of Person A and faces of Person B without much overlap. That in turn will allow you to select and remove lots of unwanted faces at once in the file manager.
Tools -> Sort -> Sort by 'Face'
- Thu Feb 03, 2022 1:01 pm
- Forum: Training Discussion
- Topic: Changing the training data after some learning
- Replies: 3
- Views: 1186
Re: Changing the training data after some learning
app.php/faqpage?sid=029d3f9703424ce2c46 ... 1b3f2#f3r4
You'd have to extract faces from the newer images and put them into the input folders that are used for training.
- Thu Jan 27, 2022 5:46 pm
- Forum: Training Discussion
- Topic: [Guide] Introducing - Phaze-A
- Replies: 89
- Views: 188066
Re: [Guide] Introducing - Phaze-A
Does it have advantages to run a model using a Phaze-A preset instead of running it without Phaze-A? In example you could choose to run H128 using the preset instead of independently. I like to think that this framework might be easier to maintain and that all presets benefit from improvements and f...
- Wed Oct 14, 2020 6:24 pm
- Forum: Training Discussion
- Topic: The realface model
- Replies: 3
- Views: 4803
The realface model
Hi. When searching for it, I found that the realface model is relatively rarely mentioned in the forum. Can somebody eleborate on how it compares to the others in training time and quality? I have seen (subjective) rankings such as dlight > dfaker == villain > original, so I wonder how it would comp...
- Wed Oct 14, 2020 6:07 pm
- Forum: Training Support
- Topic: Session Stats no longer appearing after a few hours of training
- Replies: 15
- Views: 11676
Re: Log and graph weirdness
This is usually (also) the case when the UI freezes and I let the model train some more time, before I shut the program down.
The model wouldn't be updated, but the state file would. So from then on the additional time and iterations, lost in the model, will be carried over to following sessions.