add and SE-add

[zeroboy123]

hi!，Dear author： I see two modules in the Model，add and SE-add.I did SE-add,in Reasoning process,I'm using the same RGB and a different depth as the input,I get the same result, why isn't my depth much affected，Did I input the model in the wrong way？

[danielS91]

In our experiments, SE-add leads to better results as the features get weighted using and Squeeze-and-Excitation operation before fusing them. Depending on the input, the model can decide which features to emphasize or suppress. Changing this fusion strategy after training for inference/reasoning will break the model. Moreover, all the models we trained make use of complementary information of the depth branch. So, skipping the depth input or presenting another depth image will break the model as well.