Thangka Yidam classification based on DenseNet and SENet

Xue, Panpan; Hu, Wenjin; Yue, Changtao; Tang, Huiyuan; Song, Huafei; He, Guoyuan

doi:10.1117/1.jei.31.4.043039

Cited by 5 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Currently, most scholars primarily focus on researching Thangka images from the perspectives of image restoration [2,3], object detection [4,5], and headgear segmentation [6,7], and overlook the study of Thangka images from a semantic viewpoint. Existing literature indicates a scarcity of research on image description generation for Thangka im-ages.…”

Section: Introductionmentioning

confidence: 99%

WITHDRAWN: Thangka Image Caption Generation Method Combining Multi-scale and Multi-level Aggregation

YUE,

Hu,

Zhang

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

Thangka, as a distinctive form of painting in China, plays a crucial role in facilitating a more profound appreciation and comprehension of Thangka through automated image description. Considering the diverse semantic objects and varying scales present in Thangka images, as well as their distinct spatial distribution characteristics, along with the challenge of potential information loss in image key features using Transformer-based encoding layers, this paper proposes a novel approach for generating Thangka descriptions, integrating multi-scale and multi-level aggregation. The proposed method, named Multi-scale and Multi-level Aggregation (MMA), addresses these challenges and enhances the quality of Thangka image description. At the encoding stage, we employ asymmetric convolutions to enhance the spatial information-capturing capability of convolutional layers. Additionally, we utilize a pyramid pooling module to further integrate multi-scale contextual information from both global and local regions of Thangka images, resulting in feature representations that possess rich semantic information. In the decoding stage, a multi-level aggregation network is designed to aggregate features from different encoding layers, thereby improving the utilization of semantic information from higher-level encoding layers and content information from lower-level encoding layers. This effectively addresses the issue of information loss. The experimental results demonstrate that the proposed model achieves promising performance on the Thangka dataset. Compared to the NIC model, it achieves a significant improvement of 26.7% in BLEU-4 and 0.9% in METEOR, while generating descriptions with higher accuracy.

show abstract

Section: Introductionmentioning

confidence: 99%

WITHDRAWN: Thangka Image Caption Generation Method Combining Multi-scale and Multi-level Aggregation

YUE,

Hu,

Zhang

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Figure 5. The SENet Structure [37] Similar to ResNet, SENet's layer mapping principle is to build blocks. The blocks formed in the SENet design are known as SE blocks.…”

Section: Squeeze-and-excitation Network or Senetmentioning

confidence: 99%

Implementation of Face Recognition for Patient Identification Using the Transfer Learning Method

2023

View full text Add to dashboard Cite

The hospital's status as a health center requires it to ensure patient safety, decrease incidents and treat patients. Identification of the patient is the primary source of patient safety difficulties. In addition to the patient's name and number, further patient-identifying components are needed to reduce this neglect. This work provides a solution in the form of biometric authentication, namely, face recognition. The convolutional neural network (CNN) approach can enable machine facial recognition. CNN is one of the deep learning techniques used to detect and identify picture objects. In this study, facial recognition was carried out using the transfer learning technique, VGGFace2 model pretraining, and SENet 50 model architecture. The dataset was collected via one-shot learning or a single sample per individual sampling. Applying the CNN model to the patient identification system yields two distinct outcomes: patient registration and verification. Registration utilizes a minimum distance of 0.35 and matches data with the complete database, whereas patient verification has a minimum distance of 0.28 and matches only the face in question. At the time of patient registration, the accuracy was between 90% and 100%. However, at the time of patient verification, the accuracy was 100%.

show abstract

Improvement and application of UNet network for avoiding the effect of urban dense high-rise buildings and other feature shadows on water body extraction

Xie

Chen

et al. 2023

International Journal of Remote Sensing

View full text Add to dashboard Cite

Thangka Yidam classification based on DenseNet and SENet

Cited by 5 publications

References 20 publications

WITHDRAWN: Thangka Image Caption Generation Method Combining Multi-scale and Multi-level Aggregation

WITHDRAWN: Thangka Image Caption Generation Method Combining Multi-scale and Multi-level Aggregation

Implementation of Face Recognition for Patient Identification Using the Transfer Learning Method

Improvement and application of UNet network for avoiding the effect of urban dense high-rise buildings and other feature shadows on water body extraction

Contact Info

Product

Resources

About