Comments (3)
- The existing images under
./infer_images
are not used for training our model. Our method doesn't need regularization images (although using regularization images like DreamBooth may generate better results?). Therefore, the00_start_align.sh
doesn't include any regularization images. - It seems that you are using the following training setting in
./config/stable-diffusion/aigc_id.yaml
:
data:
params:
train:
target: ldm.data.face_id.FaceIdDatasetStyleGAN3
rather than:
... ...
target: ldm.data.face_id.FaceIdDatasetOneShot
The FaceIdDatasetStyleGAN3
requires using StyleGAN3 generated faces and a pikle file listing the absolute path of each image, which is not provided yet.
We'll make this setting work, release the complete files, and modify the code and documentation ASAP!
from celebbasis.
Nvm, I got it working yay!!! No need to worry about the .pickle file for /infer_images or stylegan3 now! ^_^
I don't know exactly which of these changes I made fixed the issue, and I'm too laxy to check, but I guess these show what I did to get it to work. I had the right target in the .yaml file actually, and since FaceIdDatasetOneShot inherits from FaceIdDatasetStyleGAN3
I still needed to adjust the latter child class.
I guess there is only one id because I'm just training myself?
But In the paper it's possible to train multiple people and have them appear separate/distinct in a generated photo (which most other AIGC methods can't do btw! so that's awesome)
Then, the other change I had to make was here I think? maybe not if the defaults are overwritten anyway, I'm not sure
I just added some print statements so I could better understand the logical flow of the _load_from_pickle() method
and here you can see the print statements in the output
from celebbasis.
- We add an example using StyleGAN3 synthetic faces, 71da8d9.
- If there is only one id, the
specific ids
should be[0]
instead of[1]
. Inspecific ids
,1
means the second ID, 12156a1.
from celebbasis.
Related Issues (19)
- Code release timeline? HOT 2
- 00_align_face.sh errors out due to hardcoded paths and missing weights/.pth(s) HOT 5
- Colab link? HOT 3
- very slow start_align.py unless change is made to DataLoader instance HOT 2
- What are the CPU RAM requirements? VRAM requirements? HOT 1
- Missing /taming folder in the root directory (submodule needed?) HOT 1
- device mismatch (indices are on gpu) HOT 2
- Does this compare to roop?
- RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory HOT 1
- raise MisconfigurationException(f"No {loader_name}() method defined to run Trainer.{trainer_method}.") HOT 3
- Replacing sd checkpoint with newer version HOT 2
- What is the VRAM usage of CeleBasis?
- Amazing work and easy to convert the result for webui use!!! HOT 4
- confusing about the shape of B? HOT 4
- sd-v1-4-full-ema.ckpt HOT 1
- Celebs names
- finetuning with multiple persons
- The source images and prompt for evaluation.
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from celebbasis.