[Paper Review] Deep Learning for Generic Object Detection: A Survey

Li Liu, Wanli Ouyang+5 more|arXiv (Cornell University)|Sep 6, 2018

Advanced Neural Network Applications311 references241 citations

TL;DR

A comprehensive survey of how deep learning advances have shaped generic object detection, covering frameworks, representations, proposals, context, training, and evaluation, with future directions.

ABSTRACT

Object detection, one of the most fundamental and challenging problems in computer vision, seeks to locate object instances from a large number of predefined categories in natural images. Deep learning techniques have emerged as a powerful strategy for learning feature representations directly from data and have led to remarkable breakthroughs in the field of generic object detection. Given this period of rapid evolution, the goal of this paper is to provide a comprehensive survey of the recent achievements in this field brought about by deep learning techniques. More than 300 research contributions are included in this survey, covering many aspects of generic object detection: detection frameworks, object feature representation, object proposal generation, context modeling, training strategies, and evaluation metrics. We finish the survey by identifying promising directions for future research.

Motivation & Objective

• Summarize the landscape of generic object detection with deep learning and identify key research directions.
• Provide taxonomy and high-level organization of methods based on datasets, evaluation criteria, and problem sub-tarts.
• Highlight major milestones, datasets, and training strategies that enabled progress in generic object detection.
• Discuss open challenges and future research directions in scalable, accurate, and efficient generic object detectors.

Proposed method

• Review and categorize over 300 contributions in the field since the rise of deep learning.
• Present a taxonomy focused on datasets, evaluation metrics, context modeling, and detection proposals.
• Synthesize progress from traditional handcrafted features to deep CNN-based detectors like RCNN and beyond.
• Discuss challenges in accuracy, efficiency, and scalability, and outline future research directions.

Experimental results

Research questions

• RQ1 What are the main challenges and design considerations for generic object detection with deep learning?
• RQ2 How have datasets, evaluation metrics, and detection frameworks evolved to support progress in generic object detection?
• RQ3 What are the key milestones and methodological shifts from handcrafted features to deep learning in object detection?
• RQ4 What future directions are promising for advancing generic object detection systems?

Key findings

• Deep learning dramatically improved generic object detection performance after 2012, driving rapid progress.
• A comprehensive taxonomy helps organize methods around datasets, evaluation, context modeling, and proposal strategies.
• Large-scale annotated datasets and GPUs were pivotal for training deep detectors across many object categories.
• Progress includes transitions from handcrafted features to CNN-based detectors and region-based frameworks.
• The survey identifies remaining challenges in accuracy, efficiency, and scalability and suggests directions for future research.