Introduction
The technology can already use object detectors to correctly recognize lobsters, then count and track them.
The accuracy of the model can reach 99.5%.
Movie 1 is the actual field verification result.
Location: Kenting South Bay
Filming time: 2022.03.25
(2)Coral ecosystem analysis: Automated live coral cover analysis
Based on YOLACT, a sample segmentation algorithm released in 2019, this technology can automatically classify eight substrates from the internationally recognized Reef Check method.
The classification accuracy can reach 93.1%.
Movie 2 is the actual field verification result.
Location: Kenting South Bay
Filming time: 2022.01.26 ～ 2022.01.27
(3)Coral ecosystem analysis: Indicator biological classification
This technique assesses the ecological development of a coral area by observing its biological species.
The technology can use to classify a variety of indicator fishes and invertebrates.
The classification accuracy can reach 90%.
Movie 3 is the actual field verification result.
Location: Keelung Chaojing Park
Filming time: 2022.05
團隊已於Github提出“AI-series-on-Aquatic-Creatures-Part1_Counting”之水下生物之AI計數應用技術，此技術為可針對龍蝦之物件偵測器。 相關下載檔連結：https://github.com/softcomputinglab520/AI-series-on-Aquatic-Creatures-Part1_Counting