Some questions about your code about up-detr

[dddzg]

Thanks for your attention. The paches input is generated during the pre-training in

up-detr/datasets/selfdet.py

Line 18 in be066df

def get_random_patch_from_img(img, min_pixel=8):

For UP-DETR, the model is pre-trained on ImageNet (by running main.py) and fine-tuned on COCO (by running detr_main.py).
For pre-training and fine-tuning code, we share the engine.py file.
Please check https://github.com/dddzg/up-detr#unsupervised-pre-training for more details about how to pre-train UP-DETR on ImageNet.

You can try to read the code in datasets/selfdet.py and models/updetr.py for more details about the pre-training procedure.

[dddzg]

Feel free to re-open the issue if you have any other questions.

[Huzhen757]

OK,  Thanks！ I fully understand. I want to ask the last question：Does the code support the use of coco2017 as a pre-training data set?

…

------------------ 原始邮件 ------------------ 发件人: "dddzg/up-detr" ***@***.***>; 发送时间: 2021年10月11日(星期一) 凌晨0:32 ***@***.***>; 抄送: "Hide on ***@***.******@***.***>; 主题: Re: [dddzg/up-detr] Some questions about your code (#14) Thanks for your attention. The paches input is generated during the pre-training in https://github.com/dddzg/up-detr/blob/be066dfca5e8f1f35a4331490e8239b579bbf497/datasets/selfdet.py#L18 For UP-DETR, the model is pre-trained on ImageNet (by running main.py) and fine-tuned on COCO (by running detr_main.py). For pre-training and fine-tuning code, we share the engine.py file. Please check https://github.com/dddzg/up-detr#unsupervised-pre-training for more details about how to pre-train UP-DETR on ImageNet. You can try to read the code in datasets/selfdet.py and models/updetr.py for more details about the pre-training procedure. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.

[dddzg]

No. We expect the directory structure of the pre-training to be the following:

path/to/imagenet/
  n06785654/  # caterogey directory
    n06785654_16140.JPEG # images
  n04584207/  # caterogey directory
    n04584207_14322.JPEG # images

You can adjust the code in

up-detr/datasets/selfdet.py

Line 46 in be066df

for (troot, _, files) in os.walk(root, followlinks=True):

to support coco2017 as the pre-training dataset.

[Huzhen757]

Fine，I will try to it。 Thank you very much！

…

------------------ 原始邮件 ------------------ 发件人: "dddzg/up-detr" ***@***.***>; 发送时间: 2021年10月11日(星期一) 下午2:26 ***@***.***>; 抄送: "Hide on ***@***.******@***.***>; 主题: Re: [dddzg/up-detr] Some questions about your code (#14) No. We expect the directory structure of the pre-training to be the following: path/to/imagenet/ n06785654/ # caterogey directory n06785654_16140.JPEG # images n04584207/ # caterogey directory n04584207_14322.JPEG # images You can adjust the code in https://github.com/dddzg/up-detr/blob/be066dfca5e8f1f35a4331490e8239b579bbf497/datasets/selfdet.py#L46 to support coco2017 as the pre-training dataset. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.