Neural Information Processing Systems http://nips.cc/

Neural Information Processing Systems http://nips.cc/

We would like to thank all reviewers for their time and effort writing these valuable reviews. Reviewer 3 mentioned that a performance measure with other recent methods would be beneficial. The code for this paper will be released with the camera-ready version. In the following, we focus on the questions given by Reviewer 2. The presented network does not contain fewer parameters compared to the classical B-spline method for optimization. Furthermore, it is straightforward to extend for the 3D case.

We introduce an approach for establishing dense correspondences between partial scans of human models and a complete template model. Our approach's key novelty lies in formulating dense correspondence computation as initializing and synchronizing local transformations between the scan and the template model.

The main advantage of this approach is its efficiency at inference time with comparable performance of B-spline based approach where an optimization is needed per registration. And it has, according to the authors, much less parameters to optimize. Please confirm if this understanding is correct? 2. What is the reason of making the choice of using multiple steps to gradually transform the moving image to the fixed one? Could the local transformation done in one step instead? For instance, the position network could directly predict K locations to transform in one step instead of prediction one location for K steps.

Identifying causal relations is crucial for a variety of downstream tasks. In additional to observational data, background knowledge (BK), which could be attained from human expertise or experiments, is usually introduced for uncovering causal relations. This raises an open problem that in the presence of latent variables, what causal relations are identifiable from observational data and BK. In this paper, we propose two novel rules for incorporating BK, which offer a new perspective to the open problem. In addition, we show that these rules are applicable in some typical causality tasks, such as determining the set of possible causal effects with observational data. Our rule-based approach enhances the state-of-the-art method by circumventing a process of enumerating block sets that would otherwise take exponential complexity.

Self-supervised representation learning is heavily dependent on data augmentations to specify the invariances encoded in representations. Previous work has shown that applying diverse data augmentations is crucial to downstream performance, but augmentation techniques remain under-explored. In this work, we propose a new family of local transformations based on Gaussian random fields to generate image augmentations for self-supervised representation learning. These transformations generalize the well-established affine and color transformations (translation, rotation, color jitter, etc.) and greatly increase the space of augmentations by allowing transformation parameter values to vary from pixel to pixel. The parameters are treated as continuous functions of spatial coordinates, and modeled as independent Gaussian random fields. Empirical results show the effectiveness of the new transformations for self-supervised representation learning. Specifically, we achieve a 1.7% top-1 accuracy improvement over baseline on ImageNet downstream classification, and a 3.6% improvement on out-of-distribution iNaturalist downstream classification. However, due to the flexibility of the new transformations, learned representations are sensitive to hyperparameters. While mild transformations improve representations, we observe that strong transformations can degrade the structure of an image, indicating that balancing the diversity and strength of augmentations is important for improving generalization of learned representations.

Autonomous robots operating in real-world environments encounter a variety of objects that can be both rigid and articulated in nature. Having knowledge of these specific object properties not only helps in designing appropriate manipulation strategies but also aids in developing reliable tracking and pose estimation techniques for many robotic and vision applications. In this context, this paper presents a registration-based local region-to-region mapping approach to classify an object as either articulated or rigid. Using the point clouds of the intended object, the proposed method performs classification by estimating unique local transformations between point clouds over the observed sequence of movements of the object. The significant advantage of the proposed method is that it is a constraint-free approach that can classify any articulated object and is not limited to a specific type of articulation. Additionally, it is a model-free approach with no learning components, which means it can classify whether an object is articulated without requiring any object models or labelled data. We analyze the performance of the proposed method on two publicly available benchmark datasets with a combination of articulated and rigid objects. It is observed that the proposed method can classify articulated and rigid objects with good accuracy.