Guldan

我明白了，就是用语义二分的结果把mask=1的保留出来，再用类似原文中mean shift的聚类方法得到多个实例。

> > > bevformer代码在去年五月左右开源。 作者为什么先采用IPM生成bev特征，然后使用类似于Deter-head检测地图元素，而不是使用bevformer直接检测出地图元素呢，请问是因为效果不好，还是没有尝试过呢？ > > 您好，你尝试过BEVFormer吗？我使用BEVDepth提取BEV特征，网络训练结果并不好，loss始终不下降 您好，我最近在看Vectormapnet和MapTR的代码，想把MapTR提取BEV特征的方法用到VectormapNet上，欢迎交流qq1589841300

> I have a question about the rotation of previous bev feature. In line 152 in projects/mmdet3d_plugin/bevformer/modules/transformer.py: `tmp_prev_bev = rotate(tmp_prev_bev, rotation_angle, center=self.rotate_center)` Actually "rotation_angle" is just the yaw of prev_bev...

> @zhiqi-li @minrui-hust hello,I have a question. if I change the bev coords to [200,100](means bev_h=200,bev_w=100),and I change the rotate center to be [50,100], will it work? Thanks.

I have the same question. Have you solved the problem?

> Hi @Zhutianyi7230 , thanks for your interest! You can import rotate as BEVFormer: `from torchvision.transforms.functional import rotate` 感谢reply，len_queue=2时能跑通了。但对于代码仍存在一个问题： line138 , in projects/mmdet3d_plugin/maptr/modules/transformer.py rotation_angle = kwargs['img_metas'][i]['can_bus'][-1] 我猜想这里的本意是把prev_bev转到当前lidar坐标系下， 但是根据projects/mmdet3d_plugin/datasets/nuscenes_map_dataset.py中line1193~1202来看，kwargs['img_metas'][i]['can_bus'][-1]表示的是ego2global_rotation. 是否存在问题呢，谢谢。

> Hi,I want to ask how many days did you use to train the maptr_tiny_r50_24e model on 8-GPU setting?And I want to confirm the batch_size on each gpu is 4...