search

ldkong1205 / RoboDepth

[NeurIPS 2023] RoboDepth: Robust Out-of-Distribution Depth Estimation under Corruptions
https://ldkong.com/RoboDepth
254 stars 31 forks source link

[Competition] Question about the competition #6

Closed FrontierBreaker closed 1 year ago

FrontierBreaker commented 1 year ago
  1. Can we used pretrained weights for backbone?
  2. Can we use augmentations which are not the same as the corruption types to generate KITTI-C?

Thanks~

FrontierBreaker commented 1 year ago
  1. Can model ensemble be used?
ldkong1205 commented 1 year ago

Hi @FrontierBreaker, thanks for your interest in our competition!

Regarding your questions:

  1. Using pretrained weights for the backbone is allowed. Actually, there are some recent studies that use pre-training to improve the model's robustness. We believe this is a promising way for improving generalizability.
  2. Designing new data augmentation techniques to improve the robustness is interesting. We encourage participants to explore this direction. However, as mentioned in the terms & conditions of this competition, "any use of the 18 corruption types designed in this benchmark is strictly prohibited, including any atomic operation that is comprising any one of the mentioned corruptions". Please pay extra attention to this to avoid a possible penalty. Also, some recent works unveil that, although using a specific corruption operation as the augmentation during training might improve the performance of this corruption type, it is likely that the final scores will become lower since it tends to hurt the model's generalizability on other corruptions.
  3. Model ensemble is a common trick for most academic competitions and is also allowed in this competition.
FrontierBreaker commented 1 year ago

Thanks for your timely reply!

ArgentumHook commented 1 year ago

WARNING: Your kernel does not support swap limit capabilities or the cgroup is not mounted. Memory limited without swap. Traceback (most recent call last): File "/tmp/codalab/tmpwRT30w/run/program/evaluate.py", line 107, in pred_disps = np.load(submit_path, fix_imports=True, encoding='latin1', allow_pickle=True)["data"] File "/opt/conda/lib/python3.9/site-packages/numpy/lib/npyio.py", line 243, in getitem return format.read_array(bytes, File "/opt/conda/lib/python3.9/site-packages/numpy/lib/format.py", line 747, in read_array array = pickle.load(fp, **pickle_kwargs) ModuleNotFoundError: No module named 'torch'

Is there an abnormality in the backend runtime environment?

ldkong1205 commented 1 year ago

Hi @ArgentumHook, could you check if you have installed PyTorch into your environment? You can check this via the following command:

conda env list
ArgentumHook commented 1 year ago

Thanks for your reply, I had checked my environment and PyTorch is installed.

------------------ 原始邮件 ------------------ 发件人: "ldkong1205/RoboDepth" @.>; 发送时间: 2023年3月20日(星期一) 晚上9:13 @.>; @.**@.>; 主题: Re: [ldkong1205/RoboDepth] Question about the competition (Issue #6)

Hi @ArgentumHook, could you check if you have installed PyTorch into your environment? You can check this via the following command: conda env list
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

ldkong1205 commented 1 year ago

Hi @ArgentumHook, if so then please ensure that your environment has been activated when running the code. Let me know if you have any problems.