Mobile eye tracking captures egocentric vision and is well-suited for naturalistic studies. However, its data is noisy, especially when acquired outdoor with multiple participants over several sessions. Area of interest analysis on moving targets is difficult because A) camera and objects move nonlinearly and may disappear/reappear from the scene; and B) off-the-shelf analysis tools are limited to linearly moving objects. As a result, researchers resort to time-consuming manual annotation, which limits the use of mobile eye tracking in naturalistic studies. We introduce a method based on a fine-tuned Vision Transformer (ViT) model for classifying frames with overlaying gaze markers. After fine-tuning a model on a manually labelled training set made of 1.98% (=7845 frames) of our entire data for three epochs, our model reached 99.34% accuracy as evaluated on hold-out data. We used the method to quantify participants’ dwell time on a tablet during the outdoor user test of a mobile augmented reality application for biodiversity education. We discuss the benefits and limitations of our approach and its potential to be applied to other contexts.

Location-based augmented reality technology for real-world, outdoor experiences is rapidly gaining in popularity in a variety of fields such as engineering, education, and gaming. By anchoring medias to geographic coordinates, it is possible to design immersive experiences remotely, without necessitating an in-depth knowledge of the context. However, the creation of such experiences typically requires complex programming tools that are beyond the reach of mainstream users. We introduce BiodivAR, a web cartographic tool for the authoring of location-based AR experiences. Developed using a user-centered design methodology and open-source interoperable web technologies, it is the second iteration of an effort that started in 2016. It is designed to meet needs defined through use cases co-designed with end users and enables the creation of custom geolocated points of interest. This approach enabled substantial progress over the previous iteration. Its reliance on geolocation data to anchor augmented objects relative to the user’s position poses a set of challenges: On mobile devices, GNSS accuracy typically lies between 1 m and 30 m. Due to its impact on the anchoring, this lack of accuracy can have deleterious effects on usability. We conducted a comparative user test using the application in combination with two different geolocation data types (GNSS versus RTK). While the test’s results are undergoing analysis, we hereby present a methodology for the assessment of our system’s usability based on the use of eye-tracking devices, geolocated traces and events, and usability questionnaires.

This paper aims at underling difficulties regarding the establishment of citizen engagement processes. The specificity of citizen engagement processes lies in their evolution over time where objectives, constraints, and latitudes of a given project influence the relevance of the tools offered to citizens. Three categories of urban projects (trans-urban, major metropolitan, architectural design) have been described. These classes range from a local space with short deadlines to a regional space spread over several decades. Furthermore, the use of 3D platforms for a broad public is influenced by the users’ preferences, perception, and expertise. Throughout this study, major challenges that have been experienced during the design a 3D participatory platform are identified. They range from the issues of implementing adequate tools according to the project (temporal and spatial scalability), the participation forms (passive, consultative or interactive), to the difficulties of convincing the authorities to use new bottom-up methods. Finally, a conceptual framework for the creation of a 3D participatory platform has been introduced. It can be summarized by three major steps: (1) Meeting the needs of a decision maker, (2) Designing the participation tool in accordance with the context, (3) Translating collected raw data in order to respond to the initial request.