lllyasviel / lllyasviel/ControlNet
Abnormal image generation
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 34.1k
- Forks
- 3k
- PR merge metrics
- No merged PRs in 30d
Description
Hello, I think Controlnet is a great job, but I encountered some issues while training it.
I trained ControlNet using a set of facial datasets, and during the training process, I found that the first generated image was abnormal. Therefore, I disabled ControlNet as a trainable network, as shown in the red box above. Unfortunately, using only the frozen SD described in the paper to generate images is also abnormal.
Generate an image similar to the following image,But it should be a realistic facial image
The weight file is created using tool_ Add_ Control. py converted v1.5 version SD.
May I ask how I should solve this problem？Looking forward to your reply
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the reported conversion entry point, tool_add_control.py, and reproduce the abnormal output using the facial dataset and converted v1.5 Stable Diffusion weights. Compare generation with ControlNet trainable and frozen, then document the cause and verify that the resulting image is a realistic face.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100