search

frh23333 / mepu-owod

Code Implementation of "Unsupervised Recognition of Unknown Objects for Open-World Object Detection"
25 stars 2 forks source link

How to generate proposal file? #16

Closed vinhqng1294 closed 3 months ago

vinhqng1294 commented 7 months ago

Thank you a lot for your great work. I have a few questions about the proposal file:

  1. I have read your document, your code and I have also done some research on FreeSOLO or SoCo documents and codes to attempt to create the proposal file either by FreeSOLO or Selective Search but I cannot.
    • For FreeSOLO, I try to re-create the original proposal file the is provided, but it is not the same, the number of bboxes in the file that i generated are less than the provided one. (I follow all the steps in the scripts inference_mask.sh, then prepare and run the script gen_pseudo_labels.sh). (FreeSOLO Github link: https://github.com/NVlabs/FreeSOLO)
    • For Selective Search, I can't even know the structure of the imagenet_root (the code snippet below is in file generate_imagenet_ss_proposals.py). (For this, I reference in here https://github.com/hologerry/SoCo) 
      imagenet_root = './imagenet_root'
imagenet_root_proposals = './imagenet_root_proposals_mp'
  2. As I understand that to make an improvement for background recognition, for new classes that do not include in COCO dataset classes, I need to label new images and then add to the COCO dataset in other to enhance unknown detection, isn't it?

Please help me and correct me if I did something wrong. Thank you very much