Domain randomization-enhanced deep learning models for bird detection

Mao, Xin; Chow, Jun Kang; Tan, Pin Siang; Liu, Kuan-Fu; Wu, Jimmy; Su, Zhibin; Cheong, Ye Hur; Ooi, Ghee Leng; Pang, Chun Chiu; Wang, Yu-Hsing

doi:10.1038/s41598-020-80101-x

Cited by 8 publications

(3 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A tailor-made Green AI Camera system (Mao et al, 2021) was invented to implement automated data collection outdoors. The novel Green AI Camera system comprises six 4K-resolution cameras configured to record videos at a rate of 25 frames per second with High Efficiency Video Coding.…”

Section: Data Collectionmentioning

confidence: 99%

See 1 more Smart Citation

Artificial intelligence applications for proactive environmental monitoring and asset management

Chow

Tan

Liu

et al. 2022

Theme Issue on AI for Smart Applications

Self Cite

View full text Add to dashboard Cite

Two research studies have been implemented to explore the potential of applying artificial intelligence (AI) technologies in works projects and maintenance work of the Drainage Services Department (DSD) for enhancing the efficiency related to environmental monitoring and structural inspection, referred to as the AIEIA and AIBIM projects, respectively. In the AIEIA project, AI technologies were explored to assist observing bird behaviour that would likely be influenced by nearby DSD construction projects. A domain randomisation-enhanced model was built to detect great egrets and little egrets at Penfold Park, Hong Kong, achieving a mean average precision of 87.65%. The detection result was used to analyse the Penfold Park egretry behaviour. In the AIBIM project, AI technologies were used to facilitate the condition assessment of concrete defects in sewage treatment facilities. A classifier was developed with supervised learning for concrete defect detection, attaining recalls of 86.2% and 89.9% for the cracking and spalling classes. Another concrete defect anomaly detector was built using unsupervised learning, achieving balanced results with F2 measures of 85.2% and 76.0% for the cracking and spalling classes. The two research studies render valuable experience for the DSD to integrate AI-enabled analytics into future work to continuously improve the drainage services in Hong Kong.

show abstract

Section: Data Collectionmentioning

confidence: 99%

“…The bird size distribution was set similar to the egret size found in the recorded video. More details of deploying domain randomisation in creating synthetic egrets can be found in Mao et al (2021).…”

Section: Data Preparation Featuring the Domain Randomisation Strategymentioning

confidence: 99%

Artificial intelligence applications for proactive environmental monitoring and asset management

Chow

Tan

Liu

et al. 2022

Theme Issue on AI for Smart Applications

Self Cite

View full text Add to dashboard Cite

show abstract

“…So far, different methods have been used to detect and classify birds and bats. Many WT operators use camera-based systems to detect and classify birds using camera images and neural networks [6][7][8][9]. Here, high-resolution cameras are used to detect the animals at large distances.…”

Section: Introductionmentioning

confidence: 99%

Towards Localization and Classification of Birds and Bats in Windparks Using Multiple FMCW-Radars at Ka-Band

Zadeh¹,

Diyap²,

Moll³

et al. 2022

PIER M

View full text Add to dashboard Cite

Birds and bats are at risk when they are flying near wind turbines (WT). Hence, a protection of bats and birds is postulated to reduce their mortality, e.g., due to collisions with the rotor-blades. The use of radar technology for monitoring wind energy installations is becoming increasingly attractive for WT operators, as it offers many advantages over other sensor systems. Timely localization and classification of the approaching animal species is very crucial about the reaction measures for collision avoidance. In this work, a localization, classification, and flight path prediction technique has been developed and tested based on simulated radar signals. This allowed us to classify three different birds and one bat species with an accuracy of 90.18%. For accurate localization and target tracking, five frequency modulated continuous wave (FMCW) radars operating in Ka-Band were placed on the tower of the WT for 360 • monitoring of the WT.

show abstract

Simultaneous Localization of Multiple Defects in Software Testing Based on Reinforcement Learning

Fang

Lu²

2021

Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering

View full text Add to dashboard Cite

Domain randomization-enhanced deep learning models for bird detection

Cited by 8 publications

References 46 publications

Artificial intelligence applications for proactive environmental monitoring and asset management

Artificial intelligence applications for proactive environmental monitoring and asset management

Towards Localization and Classification of Birds and Bats in Windparks Using Multiple FMCW-Radars at Ka-Band

Simultaneous Localization of Multiple Defects in Software Testing Based on Reinforcement Learning

Contact Info

Product

Resources

About