DeepBees - Building and Scaling Convolutional Neuronal Nets For Fast and Large-Scale Visual Monitoring of Bee Hives

Marstaller, Julian; Tausch, Frederic; Stock, Simon

doi:10.1109/iccvw.2019.00036

Cited by 33 publications

(19 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Para diversificar os dados para o treinamento são necessárias prolongadas horas de gravação na colmeia, e posterior mineração desses dados com o objetivo de reconhecer ataques e inúmeras poses diferentes com e sem pólen. O próximo passo é desenvolver um trabalho de monitoramento completo de colmeias de Melipona seminigra, semelhante ao que foi feito por Marstaller et al (2019). O monitoramento deve englobar detecção de espécies invasoras nativas da Amazônia como outras espécies de abelhas, formigas, aranhas, vespas, dentre outros como forídeos.…”

Section: Resultsunclassified

“…Existem ainda sistemas com deep learning para monitoramento automatizado de colmeias de abelhas [Marstaller et al 2019]. Neste caso, as abelhas são filmadas na entrada da colmeia.…”

Section: Trabalhos Relacionadosunclassified

“…Perdas de biodiversidade entre abelhas nativas têm sido documentadas para as regiões tropicais [Sánchez-bayo e WyckhuysBorges 2019] e [Borges et al 2020]. Os principais fatores responsáveis por esse declínio incluem: infecções disseminadas por parasitas e patógenos; falta de variabilidade genética; estresse devido a movimentação sazonal de colmeias para polinização de frutas e hortaliças; resíduos de pesticidas tóxicos encontrados no pólen e néctar ou aplicado a colmeias para controlar ácaros; baixo valor nutricional de agropaisagens dominadas por monoculturas (como milho, soja e algodão); e condições climáticas mais adversas das últimas décadas [Hallmann et al 2017], [Sánchez-bayo e WyckhuysBorges 2019], [Gomes et al 2020], [Filipiak 2018] e [Marstaller et al 2019].…”

Section: Introductionunclassified

See 2 more Smart Citations

Detecção de Abelhas Nativas em Colmeias em Campo Utilizando Visão Computacional

Leocádio

Segundo

Souza

et al. 2021

Anais Do XII Workshop De Computação Aplicada À Gestão Do Meio Ambiente E Recursos Naturais (WCAMA 2021)

View full text Add to dashboard Cite

Abordagens de inteligência artificial, como visão computacional, podem ajudar a entender melhor o comportamento de abelhas e ajudar no seu manejo. Detectar espécies de abelhas em campo é ainda um desafio para os métodos tradicionais. Neste trabalho, propõe-se o uso do detector YOLO justificado por sua capacidade, velocidade e generalização. Foi possível detectar a grande maioria das abelhas presentes nos frames, com mAP de 99,5% de eficácia.

show abstract

Section: Resultsunclassified

“…Existem ainda sistemas com deep learning para monitoramento automatizado de colmeias de abelhas [Marstaller et al 2019]. Neste caso, as abelhas são filmadas na entrada da colmeia.…”

Section: Trabalhos Relacionadosunclassified

Section: Introductionunclassified

See 1 more Smart Citation

Detecção de Abelhas Nativas em Colmeias em Campo Utilizando Visão Computacional

Leocádio

Segundo

Souza

et al. 2021

Anais Do XII Workshop De Computação Aplicada À Gestão Do Meio Ambiente E Recursos Naturais (WCAMA 2021)

View full text Add to dashboard Cite

show abstract

“…The control group was fed the same amount of sugar solution during that exposure period (Gonsior et al, 2019). In a separate proof-of-concept study, the possibilities of machine learning algorithms were explored to perform localization, classification and pose estimation tasks on video recordings (Marstaller et. al., 2019).…”

Section: Fig 1 Visuelles Monitoring-system Vor Dem Eingang Des Bienementioning

confidence: 99%

Current achievements and future developments of a novel AI based visual monitoring of beehives in ecotoxicology and for the monitoring of landscape structures

Tausch

Schmidt

Diehl

2020

Preprint

Self Cite

View full text Add to dashboard Cite

in German Honigbienen sind nützliche Bio-Indikatoren. Sie besitzen ein großes Potenzial, Licht in das Ausmaß und die Wechselwirkungen der Faktoren zu bringen, die den Rückgang der Insektenzahl beeinflussen. Dieses Potenzial wurde bisher noch nicht vollständig ausgeschöpft, da die Produktion zuverlässiger Daten aufwendige und arbeitsintensive Studiendesigns erfordert, die mit hohen Kosten verbunden sind. Ein neuartiges, auf künstlicher Intelligenz (KI) basierendes visuelles Monitoringsystem könnte eine teilweise Automatisierung der Datenerfassung über Aktivität, Sammelbienenverlust und Beeinträchtigungen des zentralen Nervensystems ermöglichen. Die Möglichkeit, Merkmale aus den Bilddaten zu extrahieren, könnte darüber hinaus zukünftig auch eine Bewertung des Polleneintrags und eine Unterscheidung von toten Bienen, Drohnen und Arbeitsbienen sowie anderen Insekten wie Wespen oder Hornissen ermöglichen. Die Technologie wurde in verschiedenen Studien hinsichtlich ihrer Skalierbarkeit und ihrer Fähigkeit, bewegungs-und merkmalsbezogene Informationen zu extrahieren, validiert. Die Möglichkeiten wurden hinsichtlich ihres Potenzials analysiert, Fortschritte sowohl in der ökotoxikologischen Forschung als auch im Monitoring von Bestäuber-Lebensräumen zu ermöglichen. Abstract in EnglishHoney bees are valuable bio-indicators. As such, they hold a vast potential to help shed light on the extent and interdependencies of factors influencing the decline in the number of insects. However, to date this potential has not yet been fully leveraged, as the production of reliable data requires large-scale study designs, which are very labour intensive and therefore costly.A novel Artificial Intelligence (AI) based visual monitoring system could enable the partial automatization of data collection on activity, forager loss and impairment of the central nervous system. The possibility to extract features from image data could prospectively also allow an assessment of pollen intake and a differentiation of dead bees, drones and worker bees as well as other insects such as wasps or hornets.The technology was validated in different studies with regards to its scalability and its ability to extract motion and feature related information.The prospective possibilities were analyzed regarding their potential to enable advances both within ecotoxicological research and the monitoring of pollinator habitats.

show abstract

“…We compared (C) and further inspired our work with the help of the current state of the art in bee and insect monitoring research. In the final step of QUA 3 CK (K), we published our work at the Computer Vision for Wildlife Conservation workshop of the International Conference on Computer Vision [11].…”

Section: Pos(aisis2019)026mentioning

confidence: 99%

QUA³CK - A Machine Learning Development Process

Stock¹,

Becker²,

Grimm³

et al. 2020

Proceedings of Artificial Intelligence for Science, Industry and Society — PoS(AISIS2019)

View full text Add to dashboard Cite

Machine learning and data processing are trending topics at the moment. However, there is still a lack of a standard process to support a fast, simple, and effective development of machine learning models for academia and industry combined. Processes such as KDD or CRISP-DM are highly specialized in data mining and business cases. Therefore, engineers often refer to individual approaches to solve a machine learning problem. Especially in teaching, the lack of a standard process is a challenge. Students typically get a better understanding if a systematic approach to solve problems is given to them. A challenge when formulating a machine learning development process is to provide standard actions that work on different use-cases. At the same time, it has to be simple. Complex processes often lead to the wrong approach. The QUA 3 CK process was created at the Karlsruhe Institute of Technology to fill the gap in research and industry for a machine learning development process. However, the main focus was to reach engineering students with an easy-to-remember, didactic way to solve machine learning problems. This five-stage process starts with a machine learning question (Q), a problem that has to be solved. Understanding the data (U) comes next. Then, the loop between selecting an Algorithm (A), Adapting the features (A), and Adjusting the hyperparameters (A) is executed until the system is ready for Conclude and compare (C). At last, the Knowledge transfer (K) of the given solution can be realized as deployment in hardware or as a documentation. This paper describes the process and all individual steps in detail. Besides, we present several use-cases of QUA 3 CK in academia and research projects.

show abstract

DeepBees - Building and Scaling Convolutional Neuronal Nets For Fast and Large-Scale Visual Monitoring of Bee Hives

Cited by 33 publications

References 23 publications

Detecção de Abelhas Nativas em Colmeias em Campo Utilizando Visão Computacional

Detecção de Abelhas Nativas em Colmeias em Campo Utilizando Visão Computacional

Current achievements and future developments of a novel AI based visual monitoring of beehives in ecotoxicology and for the monitoring of landscape structures

QUA³CK - A Machine Learning Development Process

Contact Info

Product

Resources

About