High Speed and Accuracy of Animation 3D Pose Recognition Based on an Improved Deep Convolution Neural Network

Applied Sciences, Год журнала: 2023, Номер 13(13), С. 7566 - 7566

Опубликована: Июнь 27, 2023

Pose recognition in character animations is an important avenue of research computer graphics. However, the current use traditional artificial intelligence algorithms to recognize animation gestures faces hurdles such as low accuracy and speed. Therefore, overcome above problems, this paper proposes a real-time 3D pose system, which includes both facial body poses, based on deep convolutional neural networks further designs single-purpose estimation system. First, we transformed human extracted from input image abstract data structure. Subsequently, generated required at runtime dataset. This challenges conventional concept monocular estimation, extremely difficult achieve. It can also achieve running speed resolution 384 fps. The proposed method was used identify multiple-character using multiple datasets (Microsoft COCO 2014, CMU Panoptic, Human3.6M, JTA). results indicated that improved algorithm performance by approximately 3.5% 8–10 times, respectively, significantly superior other classic algorithms. Furthermore, tested system pose-recognition datasets. attitude reach 24 fps with error 100 mm, considerably less than 2D 60 learning study yielded surprisingly performance, proving deep-learning technology for has great potential.

Язык: Английский

---

A critical analysis of image-based camera pose estimation techniques

Neurocomputing, Год журнала: 2023, Номер 570, С. 127125 - 127125

Опубликована: Дек. 13, 2023

Язык: Английский

---

Disentangled body features for clothing change person re-identification

Multimedia Tools and Applications, Год журнала: 2024, Номер 83(27), С. 69693 - 69714

Опубликована: Фев. 2, 2024

Язык: Английский

---

3DPMesh: An enhanced and novel approach for the reconstruction of 3D human meshes from a single 2D image

Computers & Graphics, Год журнала: 2024, Номер 119, С. 103894 - 103894

Опубликована: Фев. 7, 2024

Язык: Английский

---

LEAPSE: Learning Environment Affordances for 3D Human Pose and Shape Estimation

IEEE Transactions on Image Processing, Год журнала: 2024, Номер 33, С. 3285 - 3300

Опубликована: Янв. 1, 2024

We live in a 3D world where people interact with each other the environment. Learning posed humans therefore requires us to perceive and interpret these interactions. This paper proposes LEAPSE, novel method that learns salient instance affordances for estimating body from single RGB image non-parametric manner. Existing methods mostly ignore environment estimate human independently surroundings. capture influences of non-contact contact instances on as an adequate representation "environment affordances". The proposed global relationships between joints, mesh vertices, body. LEAPSE achieved state-of-the-art results 3DPW dataset many affordance instances, also demonstrated excellent performance Human3.6M dataset. further demonstrate benefit our by showing existing weak models can be significantly improved when combined module.

Язык: Английский

---

Review of models for estimating 3D human pose using deep learning

PeerJ Computer Science, Год журнала: 2025, Номер 11, С. e2574 - e2574

Опубликована: Фев. 4, 2025

Human pose estimation (HPE) is designed to detect and localize various parts of the human body represent them as a kinematic structure based on input data like images videos. Three-dimensional (3D) HPE involves determining positions articulated joints in 3D space. Given its wide-ranging applications, has become one fastest-growing areas computer vision artificial intelligence. This review highlights latest advances deep-learning-based models, addressing major challenges such accuracy, real-time performance, constraints. We assess most widely used datasets evaluation metrics, providing comparison leading algorithms terms precision computational efficiency tabular form. The identifies key applications industries healthcare, security, entertainment. Our findings suggest that while deep learning models have made significant strides, handling occlusion, estimation, generalization remain. study also outlines future research directions, offering roadmap for both new experienced researchers further develop using learning.

Язык: Английский

---

Advancements in Complex Knowledge Graph Question Answering: A Survey

Electronics, Год журнала: 2023, Номер 12(21), С. 4395 - 4395

Опубликована: Окт. 24, 2023

Complex Question Answering over Knowledge Graph (C-KGQA) seeks to solve complex questions using knowledge graphs. Currently, KGQA systems achieve great success in answering simple questions, while still present challenging issues. As a result, an increasing number of novel methods have been proposed remedy this challenge. In survey, we two mainstream categories for C-KGQA, which are divided according their use graph representation and construction, namely, metric (GM)-Based Methods neural network (GNN)-based methods. Additionally, also acknowledge the influence ChatGPT, has prompted further research into utilizing graphs as source assist questions. We introduced based on pre-trained models joint reasoning. Furthermore, compiled achievements from past three years make it easier researchers with similar interests obtain state-of-the-art research. Finally, discussed resources evaluation tackling C-KGQA tasks summarized several prospects field.

Язык: Английский

---

Deep Multi-task Convolutional Neural Networks for Efficient Classification of Face Attributes

International journal of engineering. Transactions B: Applications, Год журнала: 2023, Номер 36(11), С. 2102 - 2111

Опубликована: Янв. 1, 2023

Facial feature recognition is an important subject in computer vision with numerous applications. The human face plays a significant role social interaction and personology. Valuable information such as identity, age, gender, emotions can be revealed via facial features. purpose of this paper to present technique for detecting smile, gender from images. A multi-task deep learning (MT-DL) framework was proposed that simultaneously estimate three features the remarkable accuracy. Additionally, approach aims reduce number trainable network parameters while leveraging combination different layers increase overall conducted tests demonstrate method outperforms recent advanced techniques all accuracy criteria. Moreover, it demonstrated (MTL) capable improving by 1.55% smile task, 2.04% 3.52% age task even less available data, utilizing tasks more data. Furthermore, MTL mode estimating only about 40% compared single-task mode. evaluated on IMDB-WIKI GENKI-4K datasets produced comparable state-of-the-art methods terms detection, classification.

Язык: Английский

---

SS-MVMETRO: Semi-supervised multi-view human mesh recovery transformer

Applied Intelligence, Год журнала: 2024, Номер 54(6), С. 5027 - 5043

Опубликована: Март 1, 2024

Язык: Английский

---

Fusion model with attention mechanism for carbon-neutral sports competitions

Frontiers in Ecology and Evolution, Год журнала: 2023, Номер 11

Опубликована: Июль 3, 2023

Introduction In sports competitions, using energy-saving and emission-reduction measures is an important means to achieve the carbon neutrality goal. Methods this paper, we propose attention mechanism-based convolutional neural network (CNN) combined with gated recurrent unit (GRU) for neutral energy saving emission reduction prediction model in CNN a feedforward whose input two-dimensional matrix. The main feature of that it can handle multi-channel data, use GRU make structure simple largely reduce simple, which reduces hardware computational power time cost also better solves long dependency problem RNN networks. CNN-GRU extracts data features then optimized by mechanism. Results collects real-time emissions from events, including game times, lighting usage, air conditioning other uses deep learning algorithms predict compare competition. Discussion identifying conducive realization goal has certain reference value realizing competitions under goals.

Язык: Английский

---