Related Experiment Video
Updated: Nov 7, 2025
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
744
From Handcrafted to Deep Features for Pedestrian Detection: A Survey
Summary
This survey reviews recent advances in pedestrian detection, covering both handcrafted and deep learning methods. It highlights key trends and future research directions in this critical computer vision task.
Area of Science:
- Computer Vision
- Machine Learning
Background:
- Pedestrian detection is crucial for human-centric AI tasks but remains challenging.
- Significant progress has been made using handcrafted and deep learning features over the last decade.
Purpose of the Study:
- To provide a comprehensive survey of recent advancements in pedestrian detection.
- To review both single-spectral and multi-spectral approaches.
- To analyze trends, datasets, evaluation metrics, and future research directions.
Main Methods:
- Detailed review of handcrafted feature-based methods, emphasizing shape and spatial freedom.
- Analysis of deep feature-based approaches, including pure CNN and hybrid methods.
- Exploration of multi-spectral pedestrian detection for improved robustness.
Main Results:
- Handcrafted features with high degrees of freedom show strong performance.
- In deep learning, feature enhancement, part-awareness, and post-processing are key areas.
- Multi-spectral methods offer robustness against illumination variations.
Conclusions:
- The survey identifies key trends and challenges in pedestrian detection.
- It provides a foundation for researchers by analyzing current methods and suggesting future research avenues.
