new paper code release about depth-adapted-cnn HOT 3 OPEN

Hi, the basic code and the implementation details can already be found in the provided files. You should be easily adapt the code to any existing deep network.
If you have any specific question, do not hesitate to let me know

from depth-adapted-cnn.

Sorry, there is probably something wrong with the website. I mean the code of your new paper, Depth-Adapted CNNs for RGB-D Semantic Segmentation, which is mentioned in readme.

from depth-adapted-cnn.

Hi, for the extension paper we are based on the ESAnet ICRA21 paper. As we mentioned in the paper, we show that with a simple depth-adapted conv3x3 before feeding the image into the network can significantly boost the performance and yields better performance compared to other attention works. The technical application is similar to the demo code.

I didn't organize the ESA-adapted code yet. I will try to release the code ASAP if I have some time. Please stay tuned

from depth-adapted-cnn.