kondratevakate / 3d-deformable-convolutions Goto Github PK

Hello @kondratevakate

Thank you for the great effort. I am hoping to use your work in a detection system that am working on at the moment. Deformable convolution seems to be a good strategy for my data; however, I am still trying to understand the implementation of it since it is all new to me.

It would be really nice and helpful if you maybe add a minimal example of how to use it to build a model. Is it as simple as replacing a convolutional layer by a deformable layer?

Another question, is this implementation version dependant? Am using torch 1.7.1 on a Windows machine.

Hello !
Thank you very much for you code, it's very nice.
I've taken your 3D deformable convolution to replace the last 3D "classical" convolutions in my own network.
However, I'm facing a huge increase in memory usage. Do you face a similar issue?
Thanks !
Constance

Hello, thank you for your open-source code. I tried to change the code to the TensorFlow version for medical image registration. But it works poorly. Do you have the code of the TensorFlow version? Can I learn from you?
Looking forward to your reply! Thank you very much!

Hello, I would like to ask if this repo can be used for 3D object detection in chair covers? It is possible to change the conv in the architecture to the architecture of your repo to improve the final detection accuracy. My task is to detect cars, trucks, pedestrians, etc. on the road.

Thanks

Hello, my current architecture is to use camera and radar information to perform feature extraction through the backbone of resnet 18 and 101. Is there a way for me to use 3d-deformable-convolutions to change the layer inside to deformable-convolutions to improve the effect?
Thank you~