opendet2

opendet2 copied to clipboard

CUDA error: device-side assert triggered

1

When I run train_net.py, I get this issue after loading R-50.pkl. I want to know how to solve that. Thanks a lot. My environment: CUDA 11.1, python 3.7, torch 1.8.1

Millielele

对bias参数weight_decay的处理

2

opendet2/solver/build.py line 39: 注释和代码不符，注释中bias的weight_decay为默认值，实际代码中被设为None，导致以下错误： TypeError: add(): argument 'alpha' must be Number, not NoneType

sunxuhao

Reproducibility issue

4

Hi, Amazing work on open-set detection! I trained the model after doing the dataset separation steps you suggest, and with exact same configs. The only difference is that I used...

misraya

roi head中背景类数量占比过高

1

注意到你的代码，roi head中背景类数量过高，占比约为4000/4096，是否是roi head的proposal采样没有成功？

roywang021

请问一下21-80类的有做什么特殊处理么

1

作者您好，请问一下21-80类的预测概率在训练/测试阶段有做什么特殊处理么？

roywang021

Running command python setup.py egg_info error: Multiple top-level packages discovered in a flat-layout: ['data', 'engine', 'solver', 'config', 'modeling', 'evaluation'].

1

Running command python setup.py egg_info error: Multiple top-level packages discovered in a flat-layout: ['data', 'engine', 'solver', 'config', 'modeling', 'evaluation'].

roywang021

The number class set

1

I see that you use a classification head with dim = 82. Can I directly modify the head size to 22 to implement the current design, which I think is...

Feobi1999

请问在up loss中关于background类别的处理。

3

``` targets[:num_fg, self.num_classes-2] = gt_scores[:num_fg] * \ (1-gt_scores[:num_fg]).pow(self.alpha) targets[num_fg:, self.num_classes-1] = gt_scores[num_fg:] * \ (1-gt_scores[num_fg:]).pow(self.alpha) return self._soft_cross_entropy(mask_scores, targets.detach()) ``` 看到您在这块对于background 类好像采取了和unknown类一样的操作，可以解释一下这个的原因吗？论文中好像没有看到相关的阐述。

Feobi1999

Question about Figure 2 in paper

3

Nice job! Could you please explain the "latent feature" used in Figure 2 clearer? Is it the output of Contrastive Head(CH) OR the input of CH(output of RoI Align after...

zhongxiangzju

训练数据中的开放集类别数据处理

1

你好， 在训练数据中会存在开放集的目标，在论文中提到了这个问题，但关于这部分数据的处理没有很理解，这部分是算作背景了么，会不会影响推理时候对这一类数据的置信度，麻烦稍微解释一下这部分的处理过程； 另外，如果在训练过程中把close set中的类别均标成一类，这样对训练会有影响么？ 谢谢

ltm920716