Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

mask2former_swin-s-p4-w7-224_lsj_8x2_50e_coco.pth #34

Open
wwwjiahuan opened this issue Jan 8, 2025 · 7 comments
Open

mask2former_swin-s-p4-w7-224_lsj_8x2_50e_coco.pth #34

wwwjiahuan opened this issue Jan 8, 2025 · 7 comments

Comments

@wwwjiahuan
Copy link

Where can I download mask2former_swin-s-p4-w7-224_lsj_8x2_50e_coco.pth?

@wwwjiahuan wwwjiahuan reopened this Jan 8, 2025
@JeyesHan
Copy link
Collaborator

JeyesHan commented Jan 8, 2025

@wwwjiahuan
Thanks for your attention to Infinity.
You could refer to this for downloading mask2former.
Hope you enjoy playing with Infinity!

@wwwjiahuan
Copy link
Author

Thanks for your reply. When using this weight to detect the generated images (normal), I am confused that the output always looked like this.
Summary

Total images: 1420
Total prompts: 355
% correct images: 1.13%
% correct prompts: 1.13%

Task breakdown

color_attr = 0.00% (0 / 400)
counting = 10.00% (12 / 120)
colors = 0.00% (0 / 268)
position = 0.00% (0 / 344)
two_object = 0.00% (0 / 248)
single_object = 10.00% (4 / 40)

Overall score (avg. over tasks): 0.03333

@JeyesHan
Copy link
Collaborator

JeyesHan commented Jan 9, 2025

Can you check if the generated images are normal?

@wwwjiahuan
Copy link
Author

They are normal.

@wwwjiahuan
Copy link
Author

00002
For example, the generated image is normal, but the detections in det.jsonl are abnormal.
{"filename":"output/infinity_2b_evaluation/gen_eval_cfg4_tau1_cfg_insertion_layer0_rewrite_prompt0_round2_real_rewrite/images/00294/samples/00002.jpg","tag":"color_attr","prompt":"a photo of a purple train and a yellow bicycle","correct":false,"reason":"expected train>=1, found 0\nexpected yellow bicycle>=1, found 0 yellow; and 1 purple","metadata":"{"tag": "color_attr", "include": [{"class": "train", "count": 1, "color": "purple"}, {"class": "bicycle", "count": 1, "color": "yellow"}], "prompt": "a photo of a purple train and a yellow bicycle"}","details":"{"cat": [[421.0, 634.0, 870.0, 964.0, 0.9570325016975403]], "bicycle": [[0.0, 218.0, 1024.0, 809.0, 0.9742059707641602]]}"}

@JeyesHan
Copy link
Collaborator

JeyesHan commented Jan 9, 2025

Can you run the benchmark in GenEval and get normal results? I doubt that you did not load the correct weights for mask2former.

@l-dawei
Copy link

l-dawei commented Jan 9, 2025

@wwwjiahuan
Hello, I would like to ask if you have retrained or fine-tuned Infinity?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants