zyh321

> Tty this may help you. "model.load_state_dict({k.replace('module.', ''): v for k, v in ckpt['model'].items()})" BTW, do you know why the loss value of the validation set does not seem to...

> Hi, I tried with ONNX export, but, there is some issue while inference with ONNX model because, the inference speed becomes 10 Sec per image with onnx model!!!! Hi...

@wyf0414 hi , I had this problem too, could you please tell me if you've solved it now? Thanks.

@kentang-mit hi, when I use the config file : configs/nuscenes/det/transfusion/secfpn/lidar/pointpillars.yaml to get the pointpillars backbone model. There us a part of the code that confuses me. `@BACKBONES.register_module() class PointPillarsScatter(nn.Module): def...

你好， 请问你弄明白这个问题了吗？ 应该不只是anchor顺序的问题， 在进入bbox_head之前， 特征的形状是: [n, c, n_x, n_y]， 我觉得是因为作者觉得 n_x 对应W, n_y对应H？ 所以做了transpose. 我不理解的是， 从雷达坐标系上， x正方向是向前的， y正方向是向左的， n_x就正好对应H， n_y就正好对应W 为什么要改呢

> When I test, there is a error "ModuleNotFoundError: No module named 'mmcv._ext'". I have no idea how to deal. hi [cong-ren](https://github.com/cong-ren), Did you resolve this problem? I also encountered...