Abstract:Over the past decade, deep learning models have exhibited considerable advancements, reaching or even exceeding human-level performance in a range of visual perception tasks. This remarkable progress has sparked interest in applying deep networks to real-world applications, such as autonomous vehicles, mobile devices, robotics, and edge computing. However, the challenge remains that state-of-the-art models usually demand significant computational resources, leading to impractical power consumption, latency, or carbon emissions in real-world scenarios. This trade-off between effectiveness and efficiency has catalyzed the emergence of a new research focus: computationally efficient deep learning, which strives to achieve satisfactory performance while minimizing the computational cost during inference. This review offers an extensive analysis of this rapidly evolving field by examining four key areas: 1) the development of static or dynamic light-weighted backbone models for the efficient extraction of discriminative deep representations; 2) the specialized network architectures or algorithms tailored for specific computer vision tasks; 3) the techniques employed for compressing deep learning models; and 4) the strategies for deploying efficient deep networks on hardware platforms. Additionally, we provide a systematic discussion on the critical challenges faced in this domain, such as network architecture design, training schemes, practical efficiency, and more realistic model compression approaches, as well as potential future research directions.

What problem does this paper attempt to address?

The paper primarily addresses the computational efficiency issues faced by deep learning models in the field of computer vision, aiming to solve the following core problems: 1. **Balancing Computational Efficiency and Performance**: Although deep learning models have achieved significant success in various visual tasks, even reaching or surpassing human-level performance, these models often require substantial computational resources. In practical applications (such as autonomous driving, mobile devices, robotics, etc.), this can lead to impractical power consumption, latency, or carbon emissions. 2. **Methods for Designing Efficient Deep Learning Models**: To tackle the aforementioned challenges, researchers are dedicated to developing computationally efficient deep learning models that minimize the computational cost during inference while maintaining satisfactory performance. This includes but is not limited to the following aspects: - Developing lightweight backbone networks for efficiently extracting discriminative feature representations from images, videos, or 3D scenes; - Designing specialized network architectures or algorithms for specific computer vision tasks; - Investigating deep learning model compression techniques; - Exploring strategies for deploying efficient deep networks on hardware platforms. 3. **Organization of the Review**: The paper provides an extensive and in-depth analysis of the rapidly evolving field of computationally efficient deep learning through four key areas: 1. The design of static or dynamic lightweight backbone networks for efficient and discriminative deep representation extraction; 2. Custom network architectures or algorithms tailored for specific computer vision tasks; 3. Deep learning model compression techniques; 4. Strategies for deploying efficient deep networks on hardware platforms. Through these studies, the paper aims to offer researchers in this rapidly developing field a comprehensive overview, summarize recent advancements, and highlight important challenges and future research directions.

Computation-efficient Deep Learning for Computer Vision: A Survey

Survey on Energy-Efficient Deep Neural Networks for Computer Vision

Deep Learning on Mobile and Embedded Devices: State-of-the-art, Challenges, and Future Directions

Design Automation for Fast, Lightweight, and Effective Deep Learning Models: A Survey

A Survey of Computer Vision Detection, Visual SLAM Algorithms, and Their Applications in Energy-Efficient Autonomous Systems

Efficient Deep Learning Infrastructures for Embedded Computing Systems: A Comprehensive Survey and Future Envision

Efficient Visual Recognition: A Survey on Recent Advances and Brain-inspired Methodologies

Computer Vision Model Compression Techniques for Embedded Systems: A Survey

A Survey on Green Deep Learning

Understanding Sensor Data Using Deep Learning Methods on Resource-Constrained Edge Devices.

Efficient High-Resolution Deep Learning: A Survey

Deep Model Compression for Mobile Platforms: A Survey

Model Compression for Deep Neural Networks: A Survey

Deep Learning for Edge Computing Applications: A State-of-the-Art Survey

On Efficient Training of Large-Scale Deep Learning Models: A Literature Review

Efficient Convolutional Neural Networks and Network Compression Methods for Object Detection: a Survey

A Survey On The New Generation Of Deep Learning In Image Processing

Deep Learning for Computer Vision: A Brief Review

A Survey of Deep Learning on Mobile Devices: Applications, Optimizations, Challenges, and Research Opportunities

Computer Vision and Deep Learning Transforming Image Recognition and Beyond

Deep Model Compression and Architecture Optimization for Embedded Systems: A Survey