3d reconstructions, surface visualization and skeletonization  (Thermo Fisher)

Journal: Scientific Reports

Article Title: Extracting spatial knowledge from track and field broadcasts for monocular 3D human pose estimation

doi: 10.1038/s41598-023-41142-0

Figure Lengend Snippet: Overview of common 3D human pose estimation (HPE) process. The actual real-world 3D skeleton of the athlete ( a ) is projected into image ( b ). During this step, the actual scene geometry (camera orientation and location w.r.t. athlete) ( e ) is lost. A 2D skeleton can be reliably annotated or estimated via 2D HPE (c). Recovering a 3D pose ( d ) from the 2D joint locations implies a certain geometry ( f ). Both combinations of actual pose plus geometry ( a ) &( e ) and estimated pose plus implied geometry ( d ) &( f ) result in the same 2D joint location projection ( c ) ( cf . thick blue arrows).

Article Snippet: We next compare the 2D HPE for these 50 frames to the projection of the recorded 3D Xsens skeleton using the correct scene geometry, resulting in an RMSE of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.56 \pm 3.75$$\end{document} 7.56 ± 3.75 pixel, which equals \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$50.42 \pm 28.51$$\end{document} 50.42 ± 28.51 mm.

Techniques:

Journal: Scientific Reports

Article Title: Extracting spatial knowledge from track and field broadcasts for monocular 3D human pose estimation

doi: 10.1038/s41598-023-41142-0

Figure Lengend Snippet: Multiple plausible camera parameters for a single frame. The color-graded curve shows all possible parameters, and the resulting relative camera location, consistent with the lane vanishing point. Four of these parameter sets and their 2D/3D rendering are shown (color-matched). Left: 2D Projection of scene with lanes (black) and skeleton. Center: 3D scene with 3D skeleton located consistent with the 2D Projection. Right: Overlay of all 2D projected skeletons for the respective geometries. Correct skeleton from 2D HPE in dashed black. Best viewed in color .

Article Snippet: We next compare the 2D HPE for these 50 frames to the projection of the recorded 3D Xsens skeleton using the correct scene geometry, resulting in an RMSE of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.56 \pm 3.75$$\end{document} 7.56 ± 3.75 pixel, which equals \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$50.42 \pm 28.51$$\end{document} 50.42 ± 28.51 mm.

Techniques:

Journal: Scientific Reports

Article Title: Extracting spatial knowledge from track and field broadcasts for monocular 3D human pose estimation

doi: 10.1038/s41598-023-41142-0

Figure Lengend Snippet: Comparison of the expected errors for different state-of-the-art monocular 3D HPE methods. See text for details. Error analysis for 355 frames for 16 athletes over 5 different venues and distances. Mean (Standard Deviation) .

Article Snippet: We next compare the 2D HPE for these 50 frames to the projection of the recorded 3D Xsens skeleton using the correct scene geometry, resulting in an RMSE of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$7.56 \pm 3.75$$\end{document} 7.56 ± 3.75 pixel, which equals \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$50.42 \pm 28.51$$\end{document} 50.42 ± 28.51 mm.

Techniques: Comparison, Standard Deviation

Journal: Breast Cancer Research : BCR

Article Title: Volumetric imaging and morphometric analysis of breast tumor angiogenesis using a new contrast-free ultrasound technique: a feasibility study

doi: 10.1186/s13058-022-01583-3

Figure Lengend Snippet: Set of q3D-HDMI processing steps after image formation and segmentation to prepare the image for quantification. a Microvasculature image of malignant breast tumor obtained from Amira 3D, b binary microvasculature image generated in MATLAB, c skeleton showing the network of vessels, d the skeleton overlaid on the binary vessel image. HDMI = high-definition microvasculature imaging

Article Snippet: Figure c shows the 3D skeleton image example reconstructed using MATLAB.

Techniques: Generated, Imaging