top of page

Student Group

Public·39 members

Ipi Motion Capture Keygen PORTABLE 40

Crawford noted that iPi Motion Capture software proved essential on a recent episode featuring the character Dante (video game character from Devil May Cry) vs. Bayonetta (a videogame character) in which Crawford and his team needed to capture a large amount of the fight choreography and stunts.

ipi motion capture keygen 40

Abstract:Using video sequences to restore 3D human poses is of great significance in the field of motion capture. This paper proposes a novel approach to estimate 3D human action via end-to-end learning of deep convolutional neural network to calculate the parameters of the parameterized skinned multi-person linear model. The method is divided into two main stages: (1) 3D human pose estimation based on a single frame image. We use 2D/3D skeleton point constraints, human height constraints, and generative adversarial network constraints to obtain a more accurate human-body model. The model is pre-trained using open-source human pose datasets; (2) Human-body pose generation based on video streams. Combined with the correlation of video sequences, a 3D human pose recovery method based on video streams is proposed, which uses the correlation between videos to generate a smoother 3D pose. In addition, we compared the proposed 3D human pose recovery method with the commercial motion capture platform to prove the effectiveness of the proposed method. To make a contrast, we first built a motion capture platform through two Kinect (V2) devices and iPi Soft series software to obtain depth-camera video sequences and monocular-camera video sequences respectively. Then we defined several different tasks, including the speed of the movements, the position of the subject, the orientation of the subject, and the complexity of the movements. Experimental results show that our low-cost method based on RGB video data can achieve similar results to commercial motion capture platform with RGB-D video data.Keywords: 3D human pose recovery; motion capture; generative adversarial constraint; convolutional neural network

Through the comparative assessment, we found a discrepancy of 10.7 cm in the tracked locations of body parts, and a difference of 16.2 degrees in rotation angles. However, motion detection results show that the inaccuracy of an RGB-D sensor does not have a considerable effect on action recognition in the experiment.

This paper evaluates the performance of the Kinect sensor on motion capture and action recognition for construction worker monitoring. An experimental study is undertaken to compare the accuracy of a Kinect with a commercial marker-based motion capture system, VICON, which has been used as the ground truth in prior work (Dutta 2012; Stone and Skubic 2011; Fernández-Baena et al. 2012). A VICON tracks the 3D locations of reflective markers attached to body parts with multiple cameras (e.g., 6 or 8 cameras), thereby minimizing occlusions and producing accurate tracking results. Extended from our previous work (Han et al. 2012), this paper performs the error analysis based on: (1) the estimated 3D positions of body joints, (2) the recomputed 3D rotation angles at particular joints, and (3) the effect of the motion capture accuracy on motion detection. The rest of this paper is organized as follows. Background section provides a background on the Kinect sensor and its performance evaluation. Methods section demonstrates a research methodology used to compute and analyze the three types of errors for the comparative study. Experiment section describes the experimental process for the collection of motion capture datasets with both a Kinect and a VICON. Results, including the error analysis, are presented and discussed in Results and discussion section. Finally, Conclusion section summarizes the findings of this study and suggests the direction of future research.

This section summarizes the pros and cons of an RGB-D sensor (i.e., Kinect) for motion capture, and reviews previous work on the performance evaluation of a Kinect motion capture system. Based on the literature review, further research efforts required in this domain are identified.

For motion capture, performances of the Kinect can broadly be evaluated in terms of the functionalities such as the depth measured by a sensor and body part positions estimated by motion capture solutions. This section summarizes the previous work on depth measurement and discusses issues in the pose estimation assessment.

To evaluate the impact of motion tracking accuracy on action recognition, this paper adopts the action detection framework presented in our previous work (Han et al. 2013). The framework consists of the dimension reduction of high-dimensional motion data, similarity measurements between a pair of motion data, and motion classification based on the measured similarity. First, dimension reduction is needed due to the high dimensions in motion data (e.g., 78), which hinder efficient and accurate action detection. Thus, we use Kernel Principal Component Analysis (Kernel PCA) (Schölkopf et al. 1998) to map motion data onto a 3D space, and then we compare the trajectories of datasets in the low-dimensional coordinate. In this space, a trajectory represents a sequential movement of postures (i.e., actions), and actions can be recognized by comparing the temporal patterns of transformed datasets. For the pattern recognition, temporal-spatial similarity between a pair of datasets is quantitatively measured using Dynamic Time Warping (DTW) (Okada and Hasegawa 2008). In this study, DTW measures Euclidean distances between datasets by warping the datasets in a time domain so as to compare datasets, even the sizes (i.e., durations) of which are different. For the performance evaluation, thus the similarity between a motion template (i.e., one trial of action datasets) and the entirety of the data is computed over all of the frames, and the behavior (e.g., fluctuation) of measured similarities is compared to investigate the effect of motion capture systems on the detection accuracy. Eventually, we perform the action detection that recognizes actions based on similarities by observing the ones with less similarity than a threshold (i.e., a classifier learned through classification); the detection results of Kinect and VICON datasets are compared in terms of accuracy (i.e., the fraction of correctly classified actions among all sample actions), precision (i.e., the fraction of correctly detected actions among detected ones), and recall (i.e., the fraction of correctly detected actions among ones that should be detected).

To collect motion capture data, a lab experiment was conducted in the University of Michigan 3D Lab (Han et al. 2012); experimental configuration and scenes are illustrated in Figure 3. In this experiment, actions during ladder climbing were recorded and analyzed; in construction, 16% of fatalities and 24.2% of injuries were caused by falls from a ladder in 2005 (CPWR 2008). 25 trials of each action (i.e., ascending and descending) taken by 1 subject were recorded with six 4-mega-pixel VICON sensors and a Kinect sensor. In total, 3,136 and 12,544 frames were collected with the Kinect and the VICON, respectively; and the datasets were synchronized for each system to have 3,136 frames for the comparison.

In this experiment, human skeleton models of the VICON and Kinect systems were slightly different in terms of the hierarchical structures of a human body; graphical illustrations of skeleton models extracted from each system are presented in Figure 4. Thus, for the comparison, corresponding body joints between the two systems are selected to convert the two models into the same form of a skeletal model (Figure 4c), and positions of such joints, as well as their rotation angles, are computed from motion capture data. For instance, motion capture data used in this study was in the Biovision Hierarchy (BVH) format (Meredith and Maddock 2001), in which a human posture at each frame is represented only with 3D Euler rotation angles. The BVH format also defines the 3D positions of body joints (i.e., translations) in an initial pose (e.g., T-pose as shown in Figure 4). This rotation and translation information forms a transformation matrix allowing for the computation of the 3D positions of all body joints in a global coordinate system (Meredith and Maddock 2001). To re-calculate Euler rotation angles (e.g., rotations in an order of x-, y-, and z-axes in this study) with respect to the converted skeleton model, an axis-angle between two body parts is first computed, a quaternion is defined with the axis-angle and axis vector, this quaternion forms a rotation matrix, and lastly a rotation angle is computed based on the rotation matrix (Han et al. 2012). Consequently, the 3D positions and rotation angles of each body part (Figure 4c) are compared to evaluate the tracking performances of the two systems; Table 1 describes body joint IDs corresponding to body parts in Figure 4c.

To assess the performance of the Kinect as a motion capture system, we compare it with the VICON in terms of the results of: (1) 3D positions of body joints, (2) 3D rotation angles, and (3) motion detection for the datasets simultaneously collected through a lab experiment. Based on the error analysis, the applicability of the Kinect to the motion analysis of construction workers is discussed.

SH carried out the motion analysis studies, participated the sequence alignment and drafted the manuscript. MA undertook the experiments to collect data using a motion capture system. SL and FP directed the entire processes of this study, provided suggestions on each procedure in the data collection and analysis, and reviewed the manuscript. All authors read and approved the final manuscript.

With SimulScan Document Capture, a Mobility DNA ingredient, your workers can simultaneously capture barcodes, text fields, phone numbers, images, signatures and even check boxes in the time it takes to press a button, improving invoicing and order speed and more.


Welcome to the group! You can connect with other members, ge...
bottom of page