Unsupervised meta-learning aims to learn the meta knowledge from unlabeled data and rapidly adapt to novel tasks. However, existing approaches may be misled by the context-bias (e.g. background) from the training data. In this paper, we abstract the unsupervised meta-learning problem into a Structural Causal Model (SCM) and point out that such bias arises due to hidden confounders. To eliminate the confounders, we define the priors are conditionally independent, learn the relationships between priors and intervene on them with casual factorization. Furthermore, we propose Causal Meta VAE (CMVAE) that encodes the priors into latent codes in the causal space and learns their relationships simultaneously to achieve the downstream few-shot image classification task. Results on toy datasets and three benchmark datasets demonstrate that our method can remove the context-bias and it outperforms other state-of-the-art unsupervised meta-learning algorithms because of bias-removal. Code is available at https://github.com/GuodongQi/CMVAE.

A domain adaptation scenario called few-shot hypothesis adaptation (FHA) only permits access to the well-trained source model and few labeled target data. Previous works may be misled by the discriminative yet undesirable domain-specific information since they learn the spurious correlation between inputs and predictions. In this work, our goal is to achieve the robust adaptation from the causality perspective. Specifically, we abstract FHA into a Structural Causal Model (SCM) and notice that the domain-specific variables are confounders, which lead to the spurious correlation. To eliminate the negative effects, we propose a novel Causal Interventional Few-shot hypothesis Adaption (CIFA) that consists of sampling, mixing and classification modules to perform backdoor adjustment. We also calculate the optimal solution for better convergence. Experiments on the benchmarks demonstrate the effectiveness of our algorithm.

Although supervised deep stereo matching networks have made impressive achievements, the poor generalization ability caused by the domain gap prevents them from being applied to real-life scenarios. In this paper, we propose to leverage the feature of a model trained on large-scale datasets to deal with the domain shift since it has seen various styles of images. With the cosine similarity based cost volume as a bridge, the feature will be grafted to an ordinary cost aggregation module. Despite the broad-spectrum representation, such a low-level feature contains much general information which is not aimed at stereo matching. To recover more task-specific information, the grafted feature is further input into a shallow network to be transformed before calculating the cost. Extensive experiments show that the model generalization ability can be improved significantly with this broad-spectrum and task-oriented feature. Specifically, based on two well-known architectures PSMNet and GANet, our methods are superior to other robust algorithms when transferring from SceneFlow to KITTI 2015, KITTI 2012, and Middlebury. Code is available at https://github.com/SpadeLiu/Graft-PSMNet.

Abstract Step counting is not only the key component of pedometers (which is a fundamental service on smartphones), but is also closely related to a range of applications, including motion monitoring, behavior recognition, indoor positioning and navigation. Due to the limited battery capacity of current smartphones, it is of great value to reduce the energy consumption of such a popular service. Therefore, this paper focuses on the energy efficiency of step-counting algorithms. First of all, we formulate a theoretical error model based on the well-known auto-correlation coefficient step-counting (ACSC) algorithm, so as to analyze the factors affecting step-counting accuracy. And then, in light of this model and an adaptive sampling strategy, we propose a novel energy-efficient step-counting algorithm by adaptively substituting the computationally intensive auto-correlation with simple mean absolute deviation. On these grounds, an Android pedometer is implemented. Two individual experiments are carried out and verify both the theoretical error model and the proposed algorithm. It is shown that our algorithm outperforms two famous counterparts, i.e. the original ACSC algorithm and peak detection step-counting algorithm, in terms of both accuracy and energy efficiency.

Few-shot learning (FSL) attempts to learn with limited data. In this work, we perform the feature extraction in the Euclidean space and the geodesic distance metric on the Oblique Manifold (OM). Specially, for better feature extraction, we propose a non-parametric Region Self-attention with Spatial Pyramid Pooling (RSSPP), which realizes a trade-off between the generalization and the discriminative ability of the single image feature. Then, we embed the feature to OM as a point. Furthermore, we design an Oblique Distance-based Classifier (ODC) that achieves classification in the tangent spaces which better approximate OM locally by learnable tangency points. Finally, we introduce a new method for parameters initialization and a novel loss function in the transductive settings. Extensive experiments demonstrate the effectiveness of our algorithm and it outperforms state-of-the-art methods on the popular benchmarks: mini-ImageNet, tiered-ImageNet, and Caltech-UCSD Birds-200-2011 (CUB).

Pedestrian dead reckoning (PDR) is a promising complementary technique to balance the requirements on both accuracy and costs in outdoor and indoor positioning systems. In this paper, we propose a unified framework to comprehensively tackle the three sub problems involved in PDR, including step detection and counting, heading estimation and step length estimation, based on sequentially rotating the device (reference) frame to the Earth (reference) frame through sensor fusion. To be specific, a robust step detection and counting algorithm is devised according to vertical angular velocities and turns out to be tolerant of various smartphone placements; then, a zero velocity update (ZUPT) based algorithm is leveraged to calibrate the measurements in the Earth frame; on these grounds, the heading and step length are further estimated by exploiting the cyclic features of walking. A thorough and extensive experimental analysis is conducted and confirms the effectiveness and advantages of the proposed PDR framework as well as the corresponding algorithms.