Everand Logo

Welcome to Everand!

  • Discover millions of ebooks, audiobooks, and so much more with a free trial

    From $11.99/month after trial. Cancel anytime.

    Articulated Body Pose Estimation: Unlocking Human Motion in Computer Vision
    Articulated Body Pose Estimation: Unlocking Human Motion in Computer Vision
    Articulated Body Pose Estimation: Unlocking Human Motion in Computer Vision
    Ebook114 pages1 hour

    Articulated Body Pose Estimation: Unlocking Human Motion in Computer Vision

    By Fouad Sabry

    ()

    Read preview

    About this ebook

    What is Articulated Body Pose Estimation


    In the field of computer vision, the study of techniques and systems that recover the pose of an articulated body, which is comprised of joints and rigid parts, through the use of image-based observations is referred to as the articulated body pose estimation. It is one of the longest-lasting challenges in computer vision because of the complexity of the models that relate observation with position, and because of the range of scenarios in which it would be useful.


    How you will benefit


    (I) Insights, and validations about the following topics:


    Chapter 1: Articulated body pose estimation


    Chapter 2: Image segmentation


    Chapter 3: Simultaneous localization and mapping


    Chapter 4: Gesture recognition


    Chapter 5: Video tracking


    Chapter 6: Fundamental matrix (computer vision)


    Chapter 7: Structure from motion


    Chapter 8: Bag-of-words model in computer vision


    Chapter 9: Point-set registration


    Chapter 10: Michael J. Black


    (II) Answering the public top questions about articulated body pose estimation.


    (III) Real world examples for the usage of articulated body pose estimation in many fields.


    Who this book is for


    Professionals, undergraduate and graduate students, enthusiasts, hobbyists, and those who want to go beyond basic knowledge or information for any kind of Articulated Body Pose Estimation.

    LanguageEnglish
    PublisherOne Billion Knowledgeable
    Release dateApr 29, 2024
    Articulated Body Pose Estimation: Unlocking Human Motion in Computer Vision

    Read more from Fouad Sabry

    Related to Articulated Body Pose Estimation

    Titles in the series (100)

    View More
    View More

    Related ebooks

    Intelligence (AI) & Semantics For You

    View More
    View More

    Related podcast episodes

    Related articles

    • UNLIMITED

      4D Camera Gives Robots a Wider View

      Jul 25, 2017

      Researchers have created a new camera that could create four-dimensional images and capture nearly 140 degrees of information. “We’re great at making cameras for humans but do robots need to see the way humans do? Probably not…” The camera could gene

      3 min read

    • UNLIMITED

      3d Animation Basics 03

      Apr 17, 2019

      3d Animation Basics 03
      3 min read

    • UNLIMITED

      3d Animation Basics 06

      Jul 17, 2019

      3d Animation Basics 06
      3 min read

    • UNLIMITED

      Experiments In Photogrammetry

      Jun 15, 2023

      Ever since the fire of June 30, 2021, destroyed the Lytton Museum and Archives, I have been trying to assemble preservation methods designed to reduce the effect of another catastrop loss. To this end, I have been studying ways of making digital thre

      2 min read

    • UNLIMITED

      The Eyes Have It

      Nov 9, 2021

      One of the interesting features of the new Canon EOS R3 is that it restores a feature last seen in the film EOS 3 of 1998 – that of eye-controlled autofocus. Early reviews of the camera indicate that the revived technology works well, although it req

      2 min read

    • UNLIMITED

      AI In Photography

      Jun 11, 2019

      AI In Photography
      5 min read

    • UNLIMITED

      The Backwards Evolution

      May 7, 2021

      The Backwards Evolution
      8 min read

    • UNLIMITED

      Adaptive Wide Angle Filter

      Aug 5, 2022

      Ashok Kandimalla has been in the photographic field for over three decades and has extensive experience in both film and digital photography. Being an electronics engineer by profession and a photographer, he possesses a unique and deep insight into

      6 min read

    • UNLIMITED

      For More Trustworthy AI, We May Need an ‘Interpreter’

      Jul 6, 2017

      A team of researchers is working to build trust between humans and artificial intelligence (AI) by creating an “interpreter” that can explain how an AI arrived at the answer to a specific question. In an age of self-driving cars and autonomous drones

      4 min read

    • UNLIMITED

      Professor Newman On… Specifying Autofocus

      Jul 26, 2022

      In recent articles I have detailed the way in which mirrorless autofocus differs from SLR AF, being more dependent on computational processes for its functionality. This requires that the AF system be designed around a powerful image processor. This

      2 min read

    • UNLIMITED

      Tech Lets People Play Games With Their Thoughts

      Apr 1, 2024

      Engineers have created a program that lets people use their thoughts to control video games. The innovation is part of research into brain-computer interfaces to help improve the lives of people with motor disabilities. The researchers incorporated m

      2 min read

    Related categories

    Reviews for Articulated Body Pose Estimation

    0 ratings

    0 ratings0 reviews

    What did you think?

    Tap to rate

    Review must be at least 10 words

      Book preview

      Articulated Body Pose Estimation - Fouad Sabry

      Chapter 1: Articulated body pose estimation

      The field of computer vision known as articulated body pose estimation focuses on techniques and systems that can determine the position of a moving object from a series of images of its joints and rigid parts. The difficulty of the models that relate observation with stance, as well as the wide range of applications, have made this an enduring challenge in computer vision.

      Robots need the ability to detect and understand the presence of people in their immediate vicinity. The interactive machine should comprehend the real-world context of the scenario if a human employs gestures to point to a specific object. Because of its significance and difficulty, several methods have been developed and implemented over the past two decades to address the issue of pose estimation in computer vision. Training complicated models with enormous data sets is a common approach.

      Due to the human body's 244 DOF and 230 joints, pose estimation is a challenging problem with current study. The human body has 10 major parts and 20 degrees of freedom, however not all movements between joints are visible. There is a lot of variation in look that algorithms need to account for, including variances in clothing, body shape, size, and haircuts. In addition, self-articulation occlusions, such as a person covering their face with their hand, or external occlusions can render the results unclear. Finally, most algorithms calculate pose from standard camera's monocular (two-dimensional) images. Inconsistent camera and illumination conditions may contribute to the problem. Additional performance needs only add to the complexity. These pictures have a lot of room for interpretation errors because they lack the depth information of a real body posture. Recent efforts in this direction make use of the color and depth information captured by RGBD cameras.

      In a model-based technique used by most articulated body pose estimation systems, the utmost/minimal similarity/difference between an observation (input) and a template model is used to determine an estimated pose. Various sensors, such as the following, have been considered for use in making the observation::

      Imaging at visible wavelengths, Photos taken in the long-wave infrared spectrum, Time-of-flight photography, and

      Photos taken with a laser rangefinder.

      The model makes direct use of the intermediate representations produced by these sensors. These are some of the depictions::

      Image appearance, Reconstruction based on voxels (volume elements), In three dimensions, using a total of Gaussian kernels

      Three-dimensional surface meshes.

      The human skeleton is where the concept of a part based model first emerged. When an object has the ability to articulate, it can be disassembled into component parts that can be rearranged into a variety of configurations. The primary object's scale and orientation are articulated to the pieces' scales and orientations. The springs serve to connect the many components of the model, allowing it to be described mathematically. So called because it resembles a spring, this model has other names. The compression and expansion of the springs account for the relative proximity of the various components. Spring orientations are limited by geometry. Legs, for instance, don't have arms that can rotate in a full circle. Therefore, components cannot be oriented in such a way. The number of viable combinations is therefore reduced.

      In the spring model, nodes (V) represent the components, while edges (E) represent the springs that connect them.

      Each location in the image can be reached by the x and y coordinates of the pixel location.

      Let {\displaystyle \mathbf {p} _{i}(x,\,y)} be point at {\displaystyle \mathbf {i} ^{th}} location.

      Then the cost associated in joining the spring between {\displaystyle \mathbf {i} ^{th}} and the {\displaystyle \mathbf {j} ^{th}} point can be given by {\displaystyle S(\mathbf {p} _{i},\,\mathbf {p} _{j})=S(\mathbf {p} _{i}-\mathbf {p} _{j})} .

      Hence the total cost associated in placing l components at locations {\displaystyle \mathbf {P} _{l}} is given by

      {\displaystyle S(\mathbf {P} _{l})=\displaystyle \sum _{i=1}^{l}\;\displaystyle \sum _{j=1}^{i}\;\mathbf {s} _{ij}(\mathbf {p} _{i},\,\mathbf {p} _{j})}

      The aforementioned equation is a simplification of the spring model commonly employed to describe body posture. Cost or energy function minimization is used to estimate pose from photographs. There are two terms in this energy function. The first takes into consideration how well each part matches the image data, while the second takes into account how well the oriented (deformed) portions match, such that articulation and object detection are taken into account.

      A hierarchical chain is used to build the kinematic skeleton.

      Each rigid body segment has its local coordinate system that can be transformed to the world coordinate system via a 4×4 transformation matrix {\displaystyle T_{l}} , {\displaystyle T_{l}=T_{\operatorname {par} (l)}R_{l},}

      where {\displaystyle R_{l}} denotes the local transformation from body segment S_{l} to its parent {\displaystyle \operatorname {par} (S_{l})} .

      There are three degrees of freedom (DoF) of movement at each human joint.

      Given a transformation matrix T_l , The T-pose joint position can be translated to the coordinate system of the world.

      Numerous works, the 3D joint rotation is expressed as a normalized quaternion {\displaystyle [x,y,z,w]} due to its continuity that can facilitate gradient-based optimization in the parameter estimation.

      In order to accurately estimate the poses of articulated bodies, deep learning has become the standard technique since around 2016. The appearance of the joints and the relationships between the joints of the body are learned via vast training sets as opposed to developing an explicit model for the parts as above. Extraction of 2D joint positions (keypoints), 3D joint positions, or 3D body shape from a single or several photos is typically the primary emphasis of models.

      Initial deep learning models developed were primarily concerned with determining the 2D locations of human joints from a given image. In order to detect joints, these models feed an input image into a convolutional neural network, which produces a set of heatmaps (one for each joint) with high values in those areas.

      With the proliferation of datasets containing human pose annotations from various angles,, Alongside the aforementioned research, scientists have been trying to reconstruct the 3D form of a person or animal from a collection of 2D photographs. Estimating the correct pose of the skinned multi-person linear (SMPL) model is the main focus. For each animal in the image, keypoints and a silhouette are often detected; after they are found, the parameters of a 3D shape model are typically adapted to correspond with their locations.

      Annotated photos are essential for the aforementioned algorithms, although their creation can be laborious. To solve this problem, researchers in the field of computer vision have created new algorithms that can either recognize keypoints in films without any annotations or learn 3D keypoints given only annotated 2D images from a single view.

      In the not-too-distant future, assisted living facilities may make use of personal care robots. In order for these robots

      Enjoying the preview?
      Page 1 of 1