mirror of
https://github.com/deepseek-ai/Janus.git
synced 2026-11-05 00:57:18 +00:00
[GH-ISSUE #28] Understanding the Impact of Separated Image Encoder on Training and Image Generation Quality #23
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/Janus#23
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @lavinal712 on GitHub (Nov 29, 2024).
Original GitHub issue: https://github.com/deepseek-ai/Janus/issues/28
Hello,
I am currently exploring the impact of separating the image encoder in our model architecture on both the training process and the quality of generated images. Specifically, I am interested in understanding whether this separation helps or hinders our ability to understand the images better. Additionally, I would like to discuss the potential relationship between the two aspects—whether they are mutually beneficial or if one negatively affects the other.
Key Questions:
Separation Impact: Does separating the image encoder from the rest of the model architecture help in better understanding the images during training?
Mutual Relationship: Are the training process and the quality of generated images mutually beneficial, or does one negatively impact the other?
Janus and Janusflow Quality: After conducting experiments, we observed that the quality of images generated by Janus and Janusflow is not satisfactory. Does this indicate that the method is not feasible, or are there potential improvements or adjustments that could enhance the results?
@Youngon commented on GitHub (Feb 6, 2025):
good question