Low miou when only counting the occupancy results within the field of view of the front camera

[zswzswzsw233]

Hello author, thank you for your work! Why did the MIOU decrease significantly (approximately 5 points：32->27) when only counting the occupancy results within the field of view of the front camera? I set my own camera mask to ignore the rest of the space:

At the same time, I attempted to train using only the images from the front camera, and used the camera mask to only calculate the occupancy loss within the field of view of the front camera. The results were similar.

[drilistbox]

It's a strange but interesting phenomenon. Does the instrinic parameter of the front camera same with many other camera's ?

[zswzswzsw233]

应该是训练数据不足的原因，只使用了前置相机的图片进行训练

[layemiao]

Hello, may I ask which parts you modified to achieve training and inference using only the front camera data? I want to use my own front camera for inference. Which parts of the model should I modify? Thank you very much for your reply. @zswzswzsw233

[rlczddl]

@zswzswzsw233 修改为单目时，你有遇到这个警告吗 'warning ---> no points within the predefined bev receptive field'？我调试了两天，参数和代码view transformer计算逻辑感觉找不出问题。