Please wait ...
0% Complete
Home
/
13th International Conference on Computer and Knowledge Engineering
Distilling Knowledge from CNN-Transformer Models for Enhanced Human Action Recognition
Authors :
Hamid Ahmadabadi
1
Omid Nejati Manzari
2
Ahmad Ayatollahi
3
1- School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
2- School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
3- School of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran
Keywords :
Action Recognition،Still Image،Deep Learning،Knowledge Distillation
Abstract :
This paper presents a study on improving human action recognition through the utilization of knowledge distillation, and the combination of CNN and ViT models. The research aims to enhance the performance and efficiency of smaller student models by transferring knowledge from larger teacher models. The proposed method employs a Transformer vision network as the student model, while a convolutional network serves as the teacher model. The teacher model extracts local image features, whereas the student model focuses on global features using an attention mechanism. The Vision Transformer (ViT) architecture is introduced as a robust framework for capturing global dependencies in images. Additionally, advanced variants of ViT, namely PVT, Convit, MVIT, Swin Transformer, and Twins, are discussed, highlighting their contributions to computer vision tasks. The ConvNeXt model is introduced as a teacher model, known for its efficiency and effectiveness in computer vision. The paper presents performance results for human action recognition on the Stanford 40 dataset, comparing the accuracy and mAP of student models trained with and without knowledge distillation. The findings illustrate that the suggested approach significantly improves the accuracy and mAP when compared to training networks under regular settings. These findings emphasize the potential of combining local and global features in action recognition tasks.
Papers List
List of archived papers
Extracting structural clusters from NMF feature matrix using Cosine Similarity-Based Weighted Voting
Mehdi Rahimi - Keyhan Khamforoosh - Vafa Maihami
DFIG-WECS Renewable Integration to the Grid and Stability Improvement through Optimal Damping Controller Design
Theophilus Ebuka Odoh - Aliyu Sabo - Hossien Shahinzadeh - Noor Izzri Abdul Wahab - Farshad Ebrahimi
Compressing Deep Neural Networks Using Explainable AI
Kimia Soroush - Mohsen Raji - Behnam Ghavami
Leveraging the Power of Object Detection Models in Identifying Litter for a Significant Reduction in Environmental Pollution
Lim Zhen Xian - Ervin Gubin Moung - Jason Teo Tze Wi - Nordin Saad - Farashazillah Yahya - Tiong Lin Rui - Ali Farzamnia
SingAll: Scalable Control Flow Checking for Multi-Process Embedded Systems
Mehdi Amininasab - Ahmad Patooghy - Mahdi Fazeli
BERT transformers Multitask learning Sarcasm and Sentiment classification (BMSS)
Fatemeh Molavi - Jamshid Bagherzadeh Mohasefi
Deep Learning-Driven Beamforming Optimization for High-Performance 5G Planar Antenna Arrays
Rahman Mohammadi - Seyed Reza Razavi Pour
Leveraging Self-Supervised Models for Automatic Whispered Speech Recognition
Aref Farhadipour - Homa Asadi - Volker Dellwo
Real-Time Vehicle Detection and Classification in UAV imagery Using Improved YOLOv5
Mohammad Hossein Hamzenejadi - Hadis Mohseni
Islamic Geometric algorithms: A survey
Elham Akbari - Azam Bastanfard
more
Samin Hamayesh - Version 46.0.2