Is 2D Heatmap Representation Even Necessary for Human Pose Estimation?

The 2D heatmap representation has dominated human pose estimation for years due to its high performance. However, heatmap-based approaches have some drawbacks: 1) The performance drops dramatically in the low-resolution images, which are frequently encountered in real-world scenarios. 2) To improve the localization precision, multiple upsample layers may be needed to recover the feature map resolution from low to high, which are computationally expensive. 3) Extra coordinate refinement is usually necessary to reduce the quantization error of downscaled heatmaps. To address these issues, we propose a textbf{Sim}ple yet promising textbf{D}isentangled textbf{R}epresentation for keypoint coordinate (emph{SimDR}), reformulating human keypoint localization as a task of classification. In detail, we propose to disentangle the representation of horizontal and vertical coordinates for keypoint location, leading to a more efficient scheme without extra upsampling and refinement. Comprehensive experiments conducted over COCO dataset show that the proposed emph{heatmap-free} methods outperform emph{heatmap-based} counterparts in all tested input resolutions, especially in lower resolutions by a large margin. Code will be made publicly available at url{https://github.com/leeyegy/SimDR}.

TokenPose: Learning Keypoint Tokens for Human Pose Estimation

Human pose estimation deeply relies on visual clues and anatomical constraints between parts to locate keypoints. Most existing CNN-based methods do well in visual representation, however, lacking in the ability to explicitly learn the constraint relationships between keypoints. In this paper, we propose a novel approach based on Token representation for human Pose estimation~(TokenPose). In detail, each keypoint is explicitly embedded as a token to simultaneously learn constraint relationships and appearance cues from images. Extensive experiments show that the small and large TokenPose models are on par with state-of-the-art CNN-based counterparts while being more lightweight. Specifically, our TokenPose-S and TokenPose-L achieve $72.5$ AP and $75.8$ AP on COCO validation dataset respectively, with significant reduction in parameters ($downarrow80.6%$; $downarrow$ $56.8%$) and GFLOPs ($downarrow$ $75.3%$; $downarrow$ $24.7%$). Code is publicly available.

Superconductivity in Scandium Borocarbide with orbital hybridization

Exploration of superconductivity in light element compounds has drawn considerable attention because those materials can easily realize the high $T_{c}$ superconductivity, such as ${mathrm{LnNi}}_{2}{mathrm{B}_{2}}{mathrm{C}}$ ($T_{c}$ =17 K), ${mathrm{Mg}}{mathrm{B}}_{2}$ ($T_{c}$ =39 K), and very recently super-hydrides under pressure ($T_{c}$ =250 K). Here we report the discovery of bulk superconductivity at 7.8 K in scandium borocarbide ${mathrm{Sc}}_{20}{mathrm{B}}{mathrm{C}}_{27}$ with a tetragonal lattice which structure changes based on the compound of ${mathrm{Sc}}_{3}{mathrm{C}}_{4}$ with very little B doping. Magnetization and specific heat measurements show bulk superconductivity. An upper critical field of Hc2(0) ~ 8 T is determined. Low temperature specific-heat shows that this system is a BCS fully gapped s-wave superconductor. Electronic structure calculations demonstrate that compared with ${mathrm{Sc}}_{3}{mathrm{C}}_{4}$ there are more orbital overlap and hybridization between Sc 3d electrons and 2p electrons of C-C(B)-C fragment in ${mathrm{Sc}}_{20}{mathrm{B}}{mathrm{C}}_{27}$, which form a new electric conduction path of Sc-C(B)-Sc. Those changes influence the band structure at the Fermi level and may be the reason of superconductivity in ${mathrm{Sc}}_{20}{mathrm{B}}{mathrm{C}}_{27}$.

First order transition in trigonal structure ${textbf{Ca}}{textbf{Mn}}_{2}{textbf{P}}_{2}$

We report structural and physical properties of the single crystalline ${mathrm{Ca}}{mathrm{Mn}}_{2}{mathrm{P}}_{2}$. The X-ray diffraction(XRD) results show that ${mathrm{Ca}}{mathrm{Mn}}_{2}{mathrm{P}}_{2}$ adopts the trigonal ${mathrm{Ca}}{mathrm{Al}}_{2}{mathrm{Si}}_{2}$-type structure. Temperature dependent electrical resistivity $rho(T)$ measurements indicate an insulating ground state for ${mathrm{Ca}}{mathrm{Mn}}_{2}{mathrm{P}}_{2}$ with activation energies of 40 meV and 0.64 meV for two distinct regions, respectively. Magnetization measurements show no apparent magnetic phase transition under 400 K. Different from other ${mathrm{A}}{mathrm{Mn}}_{2}{mathrm{Pn}}_{2}$ (A = Ca, Sr, and Ba, and Pn = P, As, and Sb) compounds with the same structure, heat capacity $C_{mathrm{p}}(T)$ and $rho(T)$ reveal that ${mathrm{Ca}}{mathrm{Mn}}_{2}{mathrm{P}}_{2}$ has a first-order transition at $T$ = 69.5 K and the transition temperature shifts to high temperature upon increasing pressure. The emergence of plenty of new Raman modes below the transition, clearly suggests a change in symmetry accompanying the transition. The combination of the structural, transport, thermal and magnetic measurements, points to an unusual origin of the transition.