Disentangling 3D Prototypical Networks For Few-Shot Concept Learning
This work addresses the problem of few-shot concept learning in 3D environments for applications like object detection and visual question answering, representing an incremental advance by combining existing ideas with novel architectural biases.
The paper tackles few-shot learning in 3D scenes by developing neural architectures that disentangle RGB-D images into object shapes, styles, and background maps, resulting in classifiers for object categories, color, materials, and spatial relationships that generalize better with dramatically fewer examples than state-of-the-art methods.
We present neural architectures that disentangle RGB-D images into objects' shapes and styles and a map of the background scene, and explore their applications for few-shot 3D object detection and few-shot concept classification. Our networks incorporate architectural biases that reflect the image formation process, 3D geometry of the world scene, and shape-style interplay. They are trained end-to-end self-supervised by predicting views in static scenes, alongside a small number of 3D object boxes. Objects and scenes are represented in terms of 3D feature grids in the bottleneck of the network. We show that the proposed 3D neural representations are compositional: they can generate novel 3D scene feature maps by mixing object shapes and styles, resizing and adding the resulting object 3D feature maps over background scene feature maps. We show that classifiers for object categories, color, materials, and spatial relationships trained over the disentangled 3D feature sub-spaces generalize better with dramatically fewer examples than the current state-of-the-art, and enable a visual question answering system that uses them as its modules to generalize one-shot to novel objects in the scene.