Monitoring climate change requires the analysis of vast Earth observation datasets, which exceed human capacity. Although computer vision (CV) offers transformative potential through automated image analysis, there is a critical gap between laboratory performance and operational reliability. Models struggle with domain shifts, lack uncertainty estimates, and lack workflow integration. This paper explores CV applications in key climate-related areas such as disaster assessment, agriculture and deforestation monitoring. We trace the evolution from convolutional neural networks (CNNs) to modern foundation models, and demonstrate how architectural innovations such as self-attention mechanisms, multimodal learning, and vision-language models (VLMs) enable environmental analysis at scales and speeds that were previously impractical for operational deployment. This is achieved through a systematic examination of operational systems and datasets. We propose five research directions that could maximise climate impact: real-time global monitoring systems, geospatial foundation models, multi-modal sensor fusion, uncertainty quantification for decision support, and addressing critical domain gaps.

The ability of large language models (LLMs) to process and generate text has introduced potential for applications in information extraction (IE). While it's debated whether LLMs outperform smaller fine-tuned models for classification tasks, their strong generalization capability makes them promising for domains with limited labeled data available for fine-tuning. This advantage is particularly relevant for the emerging application of the digital product passport (DPP), where the problem space is broad but domain-specific data remains scarce. Motivated by this use case, we apply generative IE to the product domain, explicitly addressing efficiency, generalizability, and data privacy constraints. We propose a two-step validation method that integrates a PLM block into the generative IE pipeline and thereby leverages LLMs' correction capability. We discover that such a validation task enhances LLM performance, particularly on the extraction of weakly expressed, low-salience entities that appear sparsely throughout the text. For certain entities, the performance of mid-size models can even reach levels comparable to larger models, and the improvement of first-step PLM predictions also enhance the final LLM output. Nevertheless, the effects on the smallest open-source LLMs (e.g., Llama-3.2 3B) is limited. Based on the findings, we develop a demo application for product information extraction that utilizes locally deployed LLMs, targeting further adaptations to real-world DPP use cases.

Research and applications in artificial intelligence have recently shifted with the rise of large pretrained models, which deliver state-of-the-art results across numerous tasks. However, the substantial increase in parameters introduces a need for parameter-efficient training strategies. Despite significant advancements, limited research has explored parameter-efficient fine-tuning (PEFT) methods in the context of transformer-based models for instance segmentation. Addressing this gap, this study investigates the effectiveness of PEFT methods, specifically adapters and Low-Rank Adaptation (LoRA), applied to two models across four benchmark datasets. Integrating sequentially arranged adapter modules and applying LoRA to deformable attention—explored here for the first time—achieves competitive performance while fine-tuning only about 1–6% of model parameters, a marked improvement over the 40–55% required in traditional fine-tuning. Key findings indicate that using 2–3 adapters per transformer block offers an optimal balance of performance and efficiency. Furthermore, LoRA, exhibits strong parameter efficiency when applied to deformable attention, and in certain cases surpasses adapter configurations. These results show that the impact of PEFT techniques varies based on dataset complexity and model architecture, underscoring the importance of context-specific tuning. Overall, this work demonstrates the potential of PEFT to enable scalable, customizable, and computationally efficient transfer learning for instance segmentation tasks.

Selecting pretrained models for image classification often involves computationally intensive finetuning. This study addresses a research gap in the standardized evaluation of transferability scores, which could simplify model selection by ranking pretrained models without exhaustive finetuning. The motivation is to reduce the computational burden of model selection through a consistent approach that guides practitioners in balancing accuracy and efficiency across tasks. This study evaluates 14 transferability scores on 11 benchmark datasets. It includes both Convolutional Neural Network (CNN) and Vision Transformer (ViT) models and ensures consistency in experimental conditions to counter the variability in previous research. Key findings reveal significant variability in score effectiveness based on dataset characteristics (e.g., fine-grained versus coarse-grained classes) and model architectures. ViT models generally show superior transferability, especially for fine-grained datasets. While no single score is best in all cases, some scores excel in specific contexts. In addition to predictive accuracy, the study also evaluates computational efficiency and identifies scores that are suitable for resource-constrained scenarios. This research provides insights for selecting appropriate transferability scores to optimize model selection strategies to facilitate efficient deployment in practice.

The development of industrial automation has led to a transformation in manufacturing, with the advent of advanced production and measurement technologies driven by digitalisation, information, and communication applications. The integration of artificial intelligence (AI) and machine learning (ML) into industrial applications represents a pivotal shift in the evolution of advanced control systems. This article examines the development and implementation of an AI-driven control system into an automated process, with a particular focus on optimizing raw material recovery through the classification of recycling items. The system developed is a prime example of the interlocking of mechanics, electronics and AI, paving the way for a new era in recycling technology. In addition, this work contributes to research in the field of industry automation and underlines the transformative power of AI in promoting a sustainable circular economy.

With the soaring popularity of electronic gadgets, Lithium-Ion Batteries (LIB) have witnessed a remarkable surge. The inspiration behind this study arises from the urgent need to automate the identification of batteries in diverse contexts, such as electronic waste recycling facilities or security screening at airports. Ultimately, it strives to minimize health hazards associated with battery recycling by enabling more accurate sorting with minimal human involvement. In this paper, we applied double transfer learning to eight cutting-edge object detectors, unlocking the potential of X-Ray images in recognizing and categorizing electronic mobile devices (EMD) along with their embedded Lithium-Ion batteries (LIB).

Meta AI has recently released the Segment Anything Model (SAM), which demonstrates exceptional zero-shot image segmentation performance across various tasks with remarkable accuracy. Despite its inability to provide accurate segmentation across multiple research fields, SAM still serves as a valuable starting point for supporting the segmentation pipeline process, particularly for tasks that require extensive and senior skills annotations. This study aims to evaluate the generalization of SAM and fine-tuning SAM models using three waste segmentation datasets. Although they are captured from real scenes as SAM was pretrained on, these datasets present several challenges, including occlusions, deformable objects, transparency, and objects easily confused with backgrounds. In our findings, the fine-tuned SAM-ViT-H model outperforms the state-ofthe-art Zerowaste, and TACO datasets with a significant increase of +30 in IoU, and it closely approaches performance levels of TrashCan 1.0, with only a -1.44 difference. After evaluating these popular waste datasets, it became evident that fine-tuning SAM as a foundational model is a crucial step for providing better generalization for downstream waste segmentation tasks. Therefore, SAM should not be disregarded or wasted.

Many of us have defective electrical devices in storerooms or drawers. Recycling electrical and electronic waste (E-waste) allows for reusing the original materials rather than mining new ones, especially if they are considered Rare Earth Elements. It is critical to separate and sort E-waste before recycling, which can be economically viable for products containing valuable metals and materials. Identification of the device and all related information will greatly support pre-dismantling. This work proposes a pipeline and automated system to identify the device using several techniques: artificial intelligence for device classification into device type, brand and model, multi-sensors for identifying internal components like Infrared or X-ray, a web crawler for creating image datasets using keywords from search engines, and a web scraper for retrieving device specifications, like dimensions, battery type, visual data, available colours, etc.

The need for detecting and sorting batteries is drastically increasing for many applications.This study proves the potential of transfer learning in predicting whether the image contains a battery or not, the location and identifying three types of batteries, namely: prismatic, pouch, and cylindrical Lithium-Ion Batteries (LIB).Particularly, it focuses on the transfer learning method in two applications: Training a large-scale dataset to detect electronic devices using a pre-trained YOLOv5m, then using these latter trained weights to detect and classify the batteries.The precision of battery detection achieves 94%, which outperforms the pretrained YOLOv5m weights with 5%, in 22 ms inference time.

Transfer Learning methods aim to reuse previously acquired knowledge about a source task to facilitate learning of a target task. In this paper, we present a Meta Learning approach to find optimal hyperparameters for Transfer Learning processes given previously known metadata about the source task, the target task, and the pre-trained model. We collected metadata and model parameters from more than 15,000 Transfer Learning processes in a dataset, which we use to learn metamodels that predict a Transfer Learning process result in terms of accuracy on the validation sets, given prior information such as the number of epochs, learning rates, optimizers, etc. Using feedforward multilayer perceptrons (MLP), we show that and how our approach finds efficient hyperparameters for Transfer Learning for image classification.

The increasing consumption of electrical and electronic devices is alarming. Therefore, the transition from linear to circular economy becomes essential. The key solution to support this transformation is artificial intelligence. This work presents a transfer learning approach to support the recycling of electrical and electronic waste (Ewaste). We emphasize the use of transfer learning technique, particularly, to classify Ewaste. In this approach, we design a hybrid model of residual nets and inception modules that can classify features of a source domain (smartphones in our case) and leverage this knowledge to another device (electric screwdrivers, as an example). Using our model, we achieve an overall accuracy of 94.27% and 97.22%, respectively. These are comparable to the popular pre-trained models, which use similar network topologies. We use a web crawler program for collecting images from search engines to build the datasets with less efforts. We show that transfer learning is more robust and performs better than training from scratch. It avoids duplication and waste of computational resources. As a result, with the benefits of transfer learning, we can provide detailed information about the devices that need to be recycled. Ultimately, this would greatly support the overall recycling process.

Transfer learning is a machine learning technique that uses previously acquired knowledge from a source domain to enhance learning in a target domain by reusing learned weights. This technique is ubiquitous because of its great advantages in achieving high performance while saving training time, memory, and effort in network design. In this paper, we investigate how to select the best pre-trained model that meets the target domain requirements for image classification tasks. In our study, we refined the output layers and general network parameters to apply the knowledge of eleven image processing models, pre-trained on ImageNet, to five different target domain datasets. We measured the accuracy, accuracy density, training time, and model size to evaluate the pre-trained models both in training sessions in one episode and with ten episodes.

Recycling electrical and electronic devices in an automated method can reduce the negative impact on human health and the environment compared with manual dismantling. This approach suggests using multi-sensors by using RGB sensor to classify the devices using a deep learning method and infrared sensor to recognize the internal component of the device. Particularly, the sorting accuracy achieves 98.5% using the state-of-the-art EfficientNet for the tested devices, and infrared images give a significant guide to the main components of the devices. Eventually, this information can be transferred to the next stage of material processing to provide fewer risks and a smarter way to avoid toxic materials.

In this paper, we present how Artificial Intelligence (AI) could support automated smartphone recycling, hence, act as an enabler for Circular Smart Cities (CSC), where the Smart City paradigm could be linked to the Circular Economy (CE), which is a leading concept of the sustainable economy. While business and society strive to gain benefits from automation, the ongoing rapid digitalization, in turn, accelerates the mass production of Waste Electric and Electronic Equipment (WEEE), often called E-Waste. Therefore, E-Waste is the fastest growing waste stream in the world and comes up with several negative environmental and social impacts. In our research, we show an AI technique (particularly, Transfer Learning) that could become an enabler for the CSC and the CE in general and supporter of automated recycling, specifically. However, research on this topic is emerging only recently, and practical applications are lacking even more. For instance, object recognition has extensive research, whereas smartphone classification nevertheless has rare attention. Our main contribution is a Transfer Learning (TL) approach based on visual-feature extraction to classify smartphones; as a result, it supports automated smartphone recycling independently of brands and even without any ex-ante information about product designs. Our findings show that the main advantages of using TL, are reducing the size of the training-set, computation time, and significant enhancements without designing a completely new network from scratch. This may ease the automated recycling of smartphones as well as other E-Waste, hence, contribute to the development of the CE and CSC.

In this paper, we present a study on Reinforcement Learning optimization models for automatic trading, in which we focus on the effects of varying the observation time. Our Reinforcement Learning agents feature a Convolutional Neural Network (CNN) together with Long Short-Term Memory (LSTM) and act on the basis of different observation time spans. Each agent tries to maximize trading profit by buying or selling one of a number of contracts in a simulated market environment for Contracts for Difference (CfD), considering correlations between individual assets by architecture. To decide which action to take on a specific contract, an agent develops a policy which relies on an observation of the whole market for a certain period of time. We investigate whether or not there exists an optimal observation sequence length, and conclude that such a value depends on market dynamics.

Sorting a huge stream of waste accurately within a short period can be done with the support of digitalization, particularly Artificial Intelligence, instead of traditional methods. The overlap of Artificial Intelligence and Circular Economy can flourish many services in the environmental technology domain, in particular smart ewaste recycling, resulting in enabling circular smart cities. We analyse the growing need for automated ewaste recycling as an essential requirement to cope with the fast growing ewaste stream and we shed the light on the impact of Artificial Intelligence in supporting the recycling process through smart classification of devices, where the smartphone is our case study. Our study applies transfer learning as a special technique of Artificial Intelligence by finetuning the output layers of AlexNet as a pretrained model and perform the implementation on a small size dataset that contains 12 classes from 6 smartphone brands. We evaluate the performance of our model by tuning the learning rate, choosing the best optimizer, and augmenting the original dataset to avoid overfitting. We found that the optimizer of Stochastic Gradient Descent with Momentum and 3e-4 as a learning rate brings almost 98

Artificial Intelligence (AI) and Machine Learning (ML) have become increasingly important for any organization that wants to stay competitive and speed up its processes. However, while organizations can choose from a variety of machine and deep learning (DL) frameworks, it is important to remember that these frameworks serve very different purposes. Therefore, the choice of a framework adapted to your needs is a decision of the utmost importance. In this article, we present an evaluation of some of the most popular machine and deep learning frameworks developed, based on an image recognition task.

The fourth industrial revolution seeks to enhance and optimize industrial processes through digital systems. However, such systems need to meet special criteria for usability and task support, ensuring users' acceptance and safety. This paper presents an approach to support employees in heavy industries with augmented reality based indoor navigation and instruction systems. An experimental study examined two different user interface concepts (navigation path vs. navigation arrow) for augmented reality head-mounted-displays. In order to validate a prototypical augmented reality application that can be deployed in such production processes, a simulated industrial environment was created. Participants walked through the scenario and were instructed to work on representative tasks, while the wearable device offered assistance and guidance. Users' perception of the system and task performance were assessed. Results indicate a superior performance of the navigation path design, as it granted participants significantly higher perceived support in the simulated working tasks. Nevertheless, the covered distance by the participants was significantly shorter in navigation arrow condition compared to the navigation path condition. Considering that the navigation path design resulted in a higher perceived Support, renders this design approach more suitable for assisting personnel working at industrial workplaces.