mzhe

 Did you use FPN to get the baseline results in the article?（without adapt baseline for example the line that AP50=25.6)

understanding about test

1

During the test, we first get the similarity between one image and a certain (or 5) object categories every episode, and finally put together the results of multiple episodes of...

evaluation errors

1

When the test was completed and evaluation started, the program encountered the following error. I checked the config file and found that the fsod dataset does not define deveit directoy....

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...

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。

what‘s the mean of scale used in ROI pooling？ Does it means：like 8 means the feature map is 1/8 size of input image？

Inference in local instead of the http demo？

2

I want get the panoptic segmentaion results of a series of images, how can I run the inference in a local terminal?