Reading book: Grokking Deep Learning.

Keep processing experiment on BEVFormer.

GrokkingDeepLearning.pdf [English Version]

GrokkingDeepLearning.pdf [Chinese Version]

1.change the backbone of FCOS-STEREO to DLA-34 ,and try to make more improvement.

2. Based on cross-attention and efficientformer . i propose a light transformer architecture regression model.