A Robotic Auto-Focus System based on Deep Reinforcement Learning

Yu, Xiaofan; Yu, Runze; Yang, Jintao; Duan, Xiaohui

doi:10.1109/icarcv.2018.8581213

Cited by 7 publications

(4 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to demonstrate the advantage of the proposed agent system in the AF task, we report a comparison with [9] and [10], since these methods are the most relevant to the present study. By replicating the described reward functions and the observation space, we train the PPO agent with the results given in Table 1.…”

Section: Single Agentsmentioning

confidence: 99%

“…The image quality assessment (IQA) is the method implemented to assess the acquired image within the RLenvironment. The reference-based (RB) method uses a ground-truth value to compare the resulting image and the Ours Phase shift [9] TEN method [10]…”

Section: Single Agentsmentioning

confidence: 99%

“…AF time (t AF ) is measured in seconds and describes the average time for an object to get into the focus plane. In [10], the time required to find the proper Focus Measure Value is obtained in the following way. An image is acquired for every lens position with a respective calculated Focus Measure Value.…”

Section: Rewardmentioning

confidence: 99%

“…In this study, we were motivated by the recent advances in reinforcement learning (RL) that provide a framework for filling the niche [9], [10]. We aimed to see if we could train hardware in the imaging system to perform image adjustments live, and improve the performance of popular embedded scene analysis models, such as Yolo, Faster R-CNN, or RetinaNet.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

DASHA: Decentralized Autofocusing System with Hierarchical Agents

Anikina¹,

Rogov²,

Dylov³

2021

Preprint

View full text Add to dashboard Cite

State-of-the-art object detection models are frequently trained offline using available datasets, such as ImageNet: large and overly diverse data that are unbalanced and hard to cluster semantically. This kind of training drops the object detection performance should the change in illumination, in the environmental conditions (e.g., rain), or in the lens positioning (out-of-focus blur) occur. We propose a decentralized hierarchical multi-agent deep reinforcement learning approach for intelligently controlling the camera and the lens focusing settings, leading to significant improvement to the capacity of the popular detection models (YOLO, Fast R-CNN, and Retina are considered). The algorithm relies on the latent representation of the camera's stream and, thus, it is the first method to allow a completely no-reference tuning of the camera, where the system trains itself to auto-focus itself.

show abstract

Section: Single Agentsmentioning

confidence: 99%

Section: Single Agentsmentioning

confidence: 99%

Section: Rewardmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

DASHA: Decentralized Autofocusing System with Hierarchical Agents

Anikina¹,

Rogov²,

Dylov³

2021

Preprint

View full text Add to dashboard Cite

show abstract

Deep Reinforcement Learning Based System for Intraoperative Hyperspectral Video Autofocusing

Budd,

Qiu,

MacCormac

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Precision autofocus in optical microscopy with liquid lenses controlled by deep reinforcement learning

Zhang,

Fu,

Shen

et al. 2024

Microsyst Nanoeng

View full text Add to dashboard Cite

Microscopic imaging is a critical tool in scientific research, biomedical studies, and engineering applications, with an urgent need for system miniaturization and rapid, precision autofocus techniques. However, traditional microscopes and autofocus methods face hardware limitations and slow software speeds in achieving this goal. In response, this paper proposes the implementation of an adaptive Liquid Lens Microscope System utilizing Deep Reinforcement Learning-based Autofocus (DRLAF). The proposed study employs a custom-made liquid lens with a rapid zoom response, which is treated as an “agent.” Raw images are utilized as the “state”, with voltage adjustments representing the “actions.” Deep reinforcement learning is employed to learn the focusing strategy directly from captured images, achieving end-to-end autofocus. In contrast to methodologies that rely exclusively on sharpness assessment as a model’s labels or inputs, our approach involved the development of a targeted reward function, which has proven to markedly enhance the performance in microscope autofocus tasks. We explored various action group design methods and improved the microscope autofocus speed to an average of 3.15 time steps. Additionally, parallel “state” dataset lists with random sampling training are proposed which enhances the model’s adaptability to unknown samples, thereby improving its generalization capability. The experimental results demonstrate that the proposed liquid lens microscope with DRLAF exhibits high robustness, achieving a 79% increase in speed compared to traditional search algorithms, a 97.2% success rate, and enhanced generalization compared to other deep learning methods.

show abstract

A Robotic Auto-Focus System based on Deep Reinforcement Learning

Cited by 7 publications

References 19 publications

DASHA: Decentralized Autofocusing System with Hierarchical Agents

DASHA: Decentralized Autofocusing System with Hierarchical Agents

Deep Reinforcement Learning Based System for Intraoperative Hyperspectral Video Autofocusing

Precision autofocus in optical microscopy with liquid lenses controlled by deep reinforcement learning

Contact Info

Product

Resources

About