Recent advances in electronics industry resulted in numerous, amazing and cheap devices, while fluent documentation and efficient, user-friendly programming environments are available for them. Modern educational systems worldwide have exploited this dynamic by including in their didactic curricula practices reflecting this progress, under the STEM umbrella. Added to this, real-world problem solving techniques increase students' interest and prepare them for their future professional role. Such challenges and job opportunities intrinsically exist in agriculture, which is a critical sector for feeding the growing population on Earth, against natural source depletion and pandemics. In this context, this work explains how STEM - based techniques can be customized to form a characteristic bouquet of activities to better prepare students for their careers in a rapidly changing environment, resulting in an innovative skillset that makes agricultural practices more efficient and environmentally friendly. The suggested reinforcement of agricultural practices with high-end technologies would make them more successful. The paradigm being presented, with most of the focus on vocational education, has been developed and tested during the STEM4Agri Erasmus project. The positive feedback acquired from the participants, encourages the adoption of similar practices into the formal educational curricula.

Rapid growth of the electronics industry resulted in a plethora of innovative and cheap devices, while fluent documentation and user-friendly programming environments are available for them. Modern educational systems worldwide have exploited this situation by incorporating practices reflecting this progress, under the STEM umbrella. Remarkable progress has been made in secondary education, but tertiary education should also follow. In this regard, the approach being presented highlights the design and implementation steps of a system for automatic classification of eggs. The whole approach utilizes open and easy-to-find software and components, e.g., arduino-like boards, cheap electronic and electromechanical modules and recyclable materials. The prototype system has been developed by students of agricultural engineering, assisted by their professors, and tested during the STEM4Agri Erasmus project, with most of the focus on its university and vocational education exploitation perspectives. According to a first set of survey findings, this STEM activity paradigm has many and multi-perspective benefits for the participants and can be seen as a flagship case of activities that should be incorporated into the curricula of educational institutes to keep in pace with the recent technological achievements and the forthcoming job transformation necessities.

The rapid growth of the electronics industry resulted in numerous, amazing and cheap devices, while fluent documentation and user-friendly programming environments are available for them. Modern educational systems worldwide have exploited this dynamic by including in their didactic curricula innovative practices that are usually called STEM actions. Added to this, enriching educational methods with real-world problem solving techniques increases students’ interest and prepares them for their future role in the society. Apparently, such challenging problems are not missing, with the depletion of natural resources to be one of the most intense ones. In this context, promising modern technological flavors like Internet of Things (IoT) and Machine Learning (ML) can join their potential to form educationally fruitful and also practically important activities targeted at increasing the social awareness for the water misuse problem, like the ones proposed herein. These activities also encourage the deployment of low-cost appliances that, only with minor modifications, can respond to a wide variety real problems in either urban or rural environments.

Since the very early beginning of the mankind history, any great difficulty, like wars or diseases, had to be a challenge for progress and innovation, otherwise the game was lost. In this regard, the recent COVID-19 pandemic provides to the learners’ and teachers’ community a great opportunity to better adapt and enrich their educational practices. Initially, aiming to assist students of agricultural engineering to demystify the innovative technologies of their scientific area, a remotely programmed and controlled robotic arm platform for fruit-picking purposes is deployed. This is just the excuse behind which a colorful bouquet of modern and software and hardware components are glued together to provide the potential for supporting a wide range of modern engineering applications. In an era that the speed of the technological achievements makes difficult to categorize their impact in industry, society or education, the proposed approach can be classified as containing mainly mixed reality, mobile, blended and project-based learning characteristics. A first set of results indicate that the discussed platform can greatly assist the students to tackle the lack of physical presence in the laboratory/classroom providing a quite interesting alternative to full in-vitro educational practices.

Land cover is a fundamental variable of the Earth's system intimately connected with many parts of the human and physical environment. Recent advances in remote sensor technology have led to the launch of spaceborne hyperspectral remote sensing sensors, such as Hyperion. The present study is exploring the potential of Hyperion hyperspectral imagery combined with the Spectral Angle Mapper (SAM) and Support Vectors Machine (SVMs) pixel-based classifiers in obtaining land cover cartography. A typical Mediterranean setting was selected as a case study, located close to the capital of Greece. Validation of the derived thematic maps was performed on the basis of the error matrix statistics using for consistency the same set of validation points. Both classifiers produced generally reasonable results with the SVMs however significantly outperforming the SAM in both overall classification accuracy and kappa coefficient. The higher classification accuracy by SVMs was attributed principally to the classifier ability to identify an optimal separating hyperplane for classes' separation which allows a low generalization error, thus producing the best possible classes' separation. Yet, as a shortcoming of both classifiers was that none of them operates on a sub-pixel level, that potentially reduces their accuracy as a result of spectral mixing problems that can be commonly found in coarse spatial resolution imagery and at fragmented landscapes.