Human pose estimation has been greatly advanced in recent years. However, even the best-performing models are not shift equivariant. In particular, a small change in input images often results in drastic alterations in output, which are problematic especially in video applications. The prevalence of top-down approaches, which typically rely on a (non-equivariant) object detector in the first stage, exac-erbates this issue. In this paper, we first demonstrate that the biased keypoint representation and the non-equivariant network components are the two main obstacles to shift equivariant pose estimation. To address the limitation, we propose an unbiased decoding method, and redesign the necessary network components (e.g., APS-ResBlock, SSP). Extensive experiments show that our method not only produces much more stable results with shifting input, but also achieves better metrics with the ability of tolerating in-accurate detector output from the first stage. To our knowledge, this is the first work to address the problem of shift equivariance in the field of pose estimation. Our method could be easily applied to existing CNN-based pose estimation networks.

The Preconditioned Conjugate Gradient method is one of the most important solvers in linear algebra system, and is widely used in scientific and engineering computing applications. Based on the Sunway heterogeneous many-core architecture, we propose a Coupled Incomplete Cholesky and Jacobi preconditioner (CICJ). The preconditioner applies a block Jacobi method to the matrix inversion in preconditioning process, localizes matrix inversions and completely eliminates the data correlation on slave cores. It strikes a better trade-off between convergence and parallelism than other preconditioners on the Sunway heterogeneous many-core architecture. Besides, a two-level software-controlled cache is designed for sparse matrix-vector multiplication operations, which makes full use of the Sunway heterogeneous many-core architecture. We apply our CICJ method on Intel, GPU, and Sunway, and the results show great generality on all three architectures. We also conduct experiments on the underwater submarine models using the open-source framework OpenFOAM. The results show that when the matrix column size is 0.82 billion and the number of non-zero values is 59 billion, our method accelerates the whole algorithm by 8.42 times compared with the diagonal incomplete Cholesky preconditioner (DIC) and 6.5 times compared with the geometric algebraic multi-grid preconditioner (GAMG) using 133,120 processors on the Sunway system.

Intrinsic image decomposition is a fundamental problem for many computer vision applications. While recent deep learning based methods have achieved very promising results on the synthetic densely labeled datasets, the results on the real-world dataset are still far from human level performance. This is mostly because collecting dense supervision on a real-world dataset is impossible. Only a sparse set of pairwise judgement from human is often used. It's very difficult for models to learn in such settings. In this paper, we investigate the possibilities of only using reflectance images for supervision during training. In this way, the demand for labeled data is greatly reduced. In order to achieve this goal, we take a deep investigation into the reflectance images. We find that reflectance images are actually comprised of two components: the flat surfaces with low frequency information, and the boundaries with high frequency details. Then, we propose to disentangle the learning process of the two components of the reflectance images. We argue that through this procedure, the reflectance images can be better modeled, and in the meantime, the shading images, though not supervised, can also achieve decent result. Extensive experiments show that our proposed network outperforms current state-of-the-art results by a large margin on the most challenging real-world IIW dataset. We also surprisingly find that on the densely labeled datasets (MIT and MPI-Sintel), our network can also achieve state-of-the-art results on both reflectance and shading images, when we only apply supervision on the reflectance images during training.