geobao

FCOS with HRSC2016: different seeds, completly different results

3

I use the default config with backbone r50 to train FCOS model on HRSC dataset. I use trainval for training and test for testing. I don't understand why I get...

In the example notebook there is this line: `!python train_distill.py --soft-loss --KL-loss --full-output-loss --with-gt-loss` However the arguments **--soft-loss** 和 **--KL-loss** don't exist in train_distill.py. Maybe outdated?

I am trying to figure out a way to start an mlflow server in read-only mode. The use case is: I log my metrics and stuff, then I want to...

I am looking at oriented annotations for ships in DOTA v1.0. The paper of DOTA says: _To make a more detailed annotation, as illustrated Fig. [3](https://www.arxiv-vanity.com/papers/1711.10398/#S3.F3), we emphasize the importance...

I can change the color of the fluid in the tank. However I cannot change the color of the label. This leads to problems such as the following example where...

你好, I would like to know what is the meaning of this line: ```python regress_weight=dict(type='iou')), ``` It is in the config file, [here](https://github.com/lzh420202/FCOSR/blob/master/configs/fcosrbox/fcosr_rx50_32x4d_fpn_3x_dota10_single.py#L48) I find it confusing because the regression...