Computer Vision for Sight

Hu, Feng; Tang, Hao; Tsema, Aleksandr; Zhu, Zhigang

doi:10.1016/b978-0-12-813445-0.00001-0

Cited by 5 publications

(2 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(1) Criterion-based pruning resorts to preserving the most important weights/attentions by employing pre-defined criteria, e.g., the L1/L2 norm [23], or activation values [6]. (2) Training-based pruning retrains models with hand-crafted sparse regularizations [36] or resource constraints [31,32].…”

Section: Introductionmentioning

confidence: 99%

“…However, the proposed methods only address the issue of accelerating self-attention mechanisms, thus are not applicable to other types of layers. Recently, Zhu et al [36] introduce the method VTP with a sparsity regularization to identify and remove unimportant patches and heads from the vision transformers. However, VTP needs to try the thresholds manually for all layers.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

X-Pruner: eXplainable Pruning for Vision Transformers

Lei¹,

Wang²

2023

Preprint

View full text Add to dashboard Cite

Recently vision transformer models have become prominent models for a range of tasks. These models, however, usually suffer from intensive computational costs, making them impractical for deployment on edge platforms. Recent studies have proposed to prune transformers in a series of criteria, such as magnitude-based, gradient-based, and mask-based. However, previous works rely heavily on hand-crafted rules and may involve time-consuming retraining or searching. As a result, measuring weight importance in an automatic and efficient way remains an open problem. To solve this problem, we propose a novel explainable pruning framework dubbed X-Pruner, by considering the explainability of the pruning criterion. Inspired by the model explanation, we propose to assign an explainability-aware mask for each prunable unit, which measures the unit's contribution to predicting every class and is fully differentiable. Then, to preserve the most informative units, we rank all units based on the absolute sum of their explainability-aware masks and using this ranking to prune enough units to meet the target resource constraint. To verify and evaluate our method, we apply the X-Pruner on representative transformer models including the DeiT and Swin Transformer. Comprehensive simulation results demonstrate that the proposed X-Pruner outperforms the state-of-the-art black-box methods with significantly reduced computational costs and slight performance degradation.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

X-Pruner: eXplainable Pruning for Vision Transformers

Lei¹,

Wang²

2023

Preprint

View full text Add to dashboard Cite

show abstract

Development of a visual to audio and tactile substitution system for mobility and orientation of visually impaired people: a review

Masal¹,

Bhatlawande²,

Shingade³

2023

Multimed Tools Appl

View full text Add to dashboard Cite

Analysis of Navigation Assistants for Blind and Visually Impaired People: A Systematic Review

2021

View full text Add to dashboard Cite

Over the last few decades, the development in the field of navigation and routing devices has become a hindering task for the researchers to develop smart and intelligent guiding mechanism at indoor and outdoor locations for blind and visually impaired people (BVIPs). The existing research need to be analysed from a historical perception including early research on the first electronic travel aids to the use of modern artificial vision models for the navigation of BVIPs. Diverse approaches such as: e-cane or guide dog, infrared-based cane, laser based walker and many others are proposed for the navigation of BVIPs. But most of these techniques have limitations such as: infrared and ultrasonic based assistance has short range capacities for object detection. While laser based assistance can harm other people if it directly hit them on their eyes or any other part of the body. These trade-offs are critical to bring this technology in practice.To systematically assess, analyze, and identify the primary studies in this specialized field and provide an overview of the trends and empirical evidence in the proposed field. This systematic research work is performed by defining a set of relevant keywords, formulating four research questions, defining selection criteria for the articles, and synthesizing the empirical evidence in this area. Our pool of studies include 191 most relevant articles to the proposed field reported between 2011 and 2020 (a portion of 2020 is included). This systematic mapping will help the researchers, engineers, and practitioners to make more authentic decisions for finding gaps in the available navigation assistants and suggest a new and enhanced smart assistant application accordingly to ensure safety and accurate guidance of the BVIPs. This research work have several implications in particular the impact of reducing fatalities and major injuries of BVIPs. INDEX TERMS Blind and visually impaired people, healthcare, smart devices, systematic literature review.

show abstract

Computer Vision for Sight

Cited by 5 publications

References 23 publications

X-Pruner: eXplainable Pruning for Vision Transformers

X-Pruner: eXplainable Pruning for Vision Transformers

Development of a visual to audio and tactile substitution system for mobility and orientation of visually impaired people: a review

Analysis of Navigation Assistants for Blind and Visually Impaired People: A Systematic Review

Contact Info

Product

Resources

About