It is evident that large language models are not only capable of generating text and images but also of facilitating discussions on the latest research in the field of human activity recognition due to their extensive knowledge. The objective of this paper is to observe how novices in machine learning in 2024 can rapidly develop their skills with the assistance of ChatGPT. As an initial introduction, we started with a dialogue-based consultation with ChatGPT and executed a classification task using the Iris dataset on Google Colaboratory. This initial task served to deepen the novices' basic understanding of machine learning. Subsequently, the authors, as part of Team Shonan-Blue, participated in the Sussex-Huawei Locomotion-Transportation (SHL) recognition challenge (hereafter SHL recognition challenge). In this challenge, the authors developed a simple human activity recognition system without expert advice and relying solely on interactions with ChatGPT. The system was constructed by integrating four classical methods: k-Nearest Neighbor algorithm (kNN), Neural Network (NN), Random Forest (RF), and Support Vector Machine (SVM). The models were selected for their differing approaches, which were anticipated to enhance prediction accuracy when combined. For feature extraction, we utilized only the primary four features (maximum, minimum, mean, standard deviation) derived from 5-second sensor data collected by accelerometers, gyroscopes, and magnetometers. The outputs of the four models were then processed ensemble to obtain the final prediction results. This yielded an F1 score of 0.936. Although the limited challenge period may not have permitted the optimal performance, the observed growth of novices who, with the assistance of ChatGPT and without the use of GPU power, managed to develop a functional system, is encouraging for future newcomers in the field of activity recognition.

A method that uses thermal images taken by UAVs is useful for outdoor search and rescue operations of a person in distress or imminent danger. To ensure saving lives under time constraints, it is necessary to improve the dependability of this method. A machine learning algorithm is used to analyze the thermal images. This involves preparing images of a heat source to be searched for as training data and test data, training a model, and using the pre-trained model to detect objects in newly captured unseen data. Dataset used for training can be selected from datasets which are publicly available on the web open sourced, or datasets that have been acquired by a user his or herself. This research is an empirical study on the intercomparison of thermal image recognition accuracy with different datasets.

If water surfaces are left untreated, mosquito larvae can develop into adult vectors within a few weeks, posing a significant risk of infectious disease. Consequently, technology for weekly on-site detection of mosquito larvae is required to control such situations effectively. Using autonomous UAV-captured images based on water surface mapping, this study establishes a technological foundation to detect mosquito larvae using deep learning, targeting automated larval source management for vector control. To enable local entomologists to build and adapt models using free, no-code cloud services, determining the minimum viable training dataset is essential. Therefore, we built an object detection model using Google Cloud Platform's Vertex AI with an extremely limited annotated dataset comprising 10 images and 371 annotations in total, of which 8 images (300 annotations) were allocated for training, to evaluate feasibility under resource-constrained conditions. The model was tested on two external datasets: controlled indoor images (JIHS, Japan) and field photographs from natural breeding sites (Malawi). Results demonstrated that while the model achieved high Recall (75.0–100.0%) and Precision (56.3–57.1%) in controlled environments (F1 0.643–0.727), Precision dropped substantially to 9.1–13.3% in turbid field conditions due to false positives from environmental noise, such as floating particles. These findings indicate that dataset diversification incorporating diverse backgrounds is essential for real-world deployment.