Fully Convolutional Geometric Features for
Category-level Object Alignment

Qiaojun Feng
Nikolay Atanasov
Department of Electrical and Computer Engineering
Contextual Robotics Institute
University of California, San Diego
In IROS 2020

This paper focuses on pose registration of different object instances from the same category. This is required in online object mapping because object instances detected at test time usually differ from the training instances. Our approach transforms instances of the same category to a normalized canonical coordinate frame and uses metric learning to train fully convolutional geometric features. The resulting model is able to generate pairs of matching points between the instances, allowing category-level registration. Evaluation on both synthetic and real-world data shows that our method provides robust features, leading to accurate alignment of instances with different shapes.


In this work, we want to align detailed CAD model to the noisy object reconstruction, when the model and the observed instance may not be identical but belongs to the same category. For example, in the bottom line we have a black office chair with wheels, and a green arm chair with four legs. They are two different instances.


We follow a matching-based alignment mechanism.

For the feature learning, we borrow the Fully Convolutional Geometric Features (FCGF) method.

We can define a normalized canonical coordinate (NCC) for one specific category. The idea is to align different instances belonging to the same category. By converting into NCC, we can generate more matching pairs across instances without introducing new data.


Video (1 min)

IROS 2020 Presentation (12 mins)


We gratefully acknowledge support from ARL DCIST CRA W911NF17-2-0181 and ONR N00014-18-1-2828.
This webpage template was borrowed from Thai Duong.