SDFEst: Categorical Pose and Shape Estimation of Objects From RGB-D Using Signed Distance Fields

Rich geometric understanding of the world is an important component many robotic applications such as planning and manipulation. In this paper, we present a modular pipeline for pose shape estimation objects from RGB-D images given their category. The core our method generative model, which integrate with novel initialization network differentiable renderer to enable 6D single or multiple views. We investigate use discretized signed distance fields efficient representation fast analysis-by-synthesis optimization. Our framework enables multi-view optimization extensibility. demonstrate benefits approach over state-of-the-art methods in several experiments on both synthetic real data. open-source at https://github.com/roym899/sdfest .