6D pose Estimation on Jetson Nano

Hi. I am working on 6DoF object pose estimation based on RGBD recently for som household stuff, such as cola can, water bottle … I have done a lot of researches, but I didn’t find a proper methods available on Nano platform as for the computin cost.
They are usually a registration problem from key points between the rgbd observation of object and CAD model. Even for the categorical level methods, a huge memory needed for the point cloud processing.

Are there some methods that could deploy to Jetson Nano? Nowadays, I turned my concerntration from deep learning based models to classical 3D descriptors, such as SHOT and CVFH. If you have experiences which can be deployed on a Nano, please give me a brief introduction. Thanks.


Most of the deep learning frameworks(ex. PyTorch, TensorFlow) can run on the Jetson platform.
But due to the limited resources of Nano, it’s recommended to convert the model into ONNX and then deploy it with TensorRT.


This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.