books by Juergen Landauer
papers by Juergen Landauer
MDPI Heritage, 2024
The value of artificial intelligence and machine learning applications for use in heritage
resear... more The value of artificial intelligence and machine learning applications for use in heritage
research is increasingly appreciated. In specific areas, notably remote sensing, datasets have increased
in extent and resolution to the point that manual interpretation is problematic and the availability
of skilled interpreters to undertake such work is limited. Interpretation of the geophysical datasets
associated with prehistoric submerged landscapes is particularly challenging. Following the Last
Glacial Maximum, sea levels rose by 120 m globally, and vast, habitable landscapes were lost to the
sea. These landscapes were inaccessible until extensive remote sensing datasets were provided by the
offshore energy sector. In this paper, we provide the results of a research programme centred on AI
applications using data from the southern North Sea. Here, an area of c. 188,000 km2 of habitable
terrestrial land was inundated between c. 20,000 BP and 7000 BP, along with the cultural heritage
it contained. As part of this project, machine learning tools were applied to detect and interpret
features with potential archaeological significance from shallow seismic data. The output provides a
proof-of-concept model demonstrating verifiable results and the potential for a further, more complex,
leveraging of AI interpretation for the study of submarine palaeolandscapes.
CAA 2021 – “Digital Crossroads”, 2021
Nowadays, archaeologists have vast amounts of Light Detection and Ranging (LiDAR) and other remot... more Nowadays, archaeologists have vast amounts of Light Detection and Ranging (LiDAR) and other remote sensing data at their disposal, to search for previously undiscovered archaeological objects, often at a national scale. This leads to a Big Data problem in archaeology: some degree of automation is needed, as humans alone cannot cope with these ever-growing data sources. In this research, we have developed a novel workflow based on the Artificial Intelligence (AI) technology of Convolutional Neural Networks (CNNs), to automate the detection of unknown, complex archaeological objects. Our hypothesis is that a high-quality remote sensing data source such as LiDAR and a curated list of known objects, is sufficient to find a large number-or ideally all-additional undiscovered objects within a landscape. In a case study presented here, we use Prehistoric hillforts in England as an example for this workflow and present a three-step approach to demonstrate its efficiency.
Automating Archaeological Documentation with Robotics Tools Photogrammetry has been adopted by ar... more Automating Archaeological Documentation with Robotics Tools Photogrammetry has been adopted by archaeologists for documentation of artefacts for quite a while with good results. Even researchers with limited experience are able to produce good results by obeying simple rules such as “ensure sufficient overlap between images” or by using software tools such as Pix4D or Scann3D which incorporate such rules. However, these rules are mere simplifications of the underlying photogrammetry algorithms. Whereas they work well with rather planar structures and objects without protruding or partially occluded features, they are bound to provide less-than-optimal results with a large set of real world archaeological objects. Here, we propose a new approach derived from recent progress in robotics and computer vision (more specifically solutions to the so-called Next-Best-View problem) which lets the computer decide which image to take next based on previous images. By intertwining image capturing and processing a more accurate 3D documentation can be achieved. We demonstrate that this method is useful for both entire sites (demonstrated with a Roman Villa Rustica) and smaller scale objects (e.g. Minoan pottery). Capturing the latter currently still requires some manual assistance, whereas site recordings are fully automatic (with drones).
Archaeological Prospection, 2021
This paper presents new developments on drone-based automated survey for the detection of individ... more This paper presents new developments on drone-based automated survey for the detection of individual items or fragments of material culture visible on the ground surface. Since the publication of our original proof of concept, awarded with the
Journal of Cultural Heritage, 2021
Proceedings of the third symposium on Virtual reality modeling language - VRML '98, 1998
Journal of Visual Languages & Computing, 1997
Proceedings of Symposium on Visual Languages, 1995
ABSTRACT
… and Systems' 97. …, 1997
... provide additional languages such as Python [lo], OML [7], or, for high-level interaction, Li... more ... provide additional languages such as Python [lo], OML [7], or, for high-level interaction, Lisp [ll ... Lightning could not rely on a single language but had to provide means to support nearly all ... [lo] Pausch, R et al., A Brief Architectural Overview of Alice, a Rapid Prototyping System for ...
Image Segmentation To Locate Ancient Maya Architectures Using Deep Learning, 2021
Exploration of the Maya forest region remotely through machine learning has recently accelerated.... more Exploration of the Maya forest region remotely through machine learning has recently accelerated. Using experts to manually look at satellite data is time-consuming and expensive. The machine learning competition Discover the mysterious of the Maya addresses this problem and calls for a competition to improve the performance of state-of-the-art models to automatically detect objects of interest using satellite images. With a given LiDAR image, the model should detect three classes of objects: Aguadas, buildings and platforms. We have set up a pipeline that essentially consists of three steps. First, we generate synthetic data in three different ways to increase the training set. In the second step, we mix them with the real training data and then train an ensemble of DeepLabV3+ and HRnet networks. In the third step, we applied thresholds to improve the segmentation masks. We achieved an average intersection over union (IOU) of 0.8275 for all three classes and the best score of 0.7569 for the building class.
Archaeological Prospection, 2021
This paper presents new developments on drone‐based automated survey for the detection of individ... more This paper presents new developments on drone‐based automated survey for the detection of individual items or fragments of material culture visible on the ground surface. Since the publication of our original proof of concept, awarded with the Journal of Archaeological Science and Society for Archaeological Sciences Emerging Investigator Award 2019, additional funding has allowed us to implement a series of improvements to the method. These aim to improve detection capabilities and the extraction of items' shapes and increase flight autonomy, control, area covered per flight and the type of environments in which the method can be applied while reducing computing needs, processing time and expertise necessary for its application. This paper provides an account of the methods followed to achieve these objectives, their preliminary results and the current development for their implementation into a free and open‐source system that can be used by the archaeological community at large.
Future Generation Computer Systems, 1998
Virtual reality (VR) systems utilize more and more input and output media in order to make virtua... more Virtual reality (VR) systems utilize more and more input and output media in order to make virtual environments more realistic. Furthermore, VR developers nowadays concentrate on rich interaction and behavior. This causes the need for a new type of VR development system architecture. In this paper such a new architecture is proposed and discussed. Its key features include support for new input and output media, device independence, and rapid behavioral prototyping. The Lightning VR system is introduced as a prototype for the presented architecture. It has already shown its usefulness in several industrial applications as design reviews, marketing events, assembly planing, etc.
Journal of Cultural Heritage, 2020
The systematic mapping of hollow roads, traces of (post)medieval sunken cart tracks ways, can pro... more The systematic mapping of hollow roads, traces of (post)medieval sunken cart tracks ways, can provideinformation on past human movement and historical route networks. However, the sheer amount oftraces and of available high-quality data necessitates the use of computational methods for the automa-tic detection of these archaeological objects. Therefore, a novel approach, named CarcassonNet, has beendeveloped that uses a combination of a Deep Learning convolutional neural network and image proces-sing algorithms to detect and trace hollow roads in LiDAR data from the Netherlands. CarcassonNet hasbeen specifically developed for the archaeological domain, focusing on being computationally light andsuited for reconstructing partially preserved and intersected hollow roads. Instead of using the wholeroads as input for the convolutional neural network, in CarcassonNet individual sections are used. Thismakes it much more cost-effective to create a sufficient training dataset, and makes the classificationtask (performed by the neural network) relatively simple, with better detection results. The output ofCarcassonNet consists of two types of geospatial vectors that offer the opportunity to efficiently studythe roads themselves and their precise location in the landscape (polygons), and the course of the roadsand the resulting route network (lines). An experimental evaluation shows that CarcassonNet is able toeffectively detect hollow roads, with a MCC score of 0.47. Furthermore, it is shown that using the DigitalTerrain Model, instead of visualized LiDAR data (hillshade) improves the performance of the convolutionalneural network. The results of this research offer opportunities to reconstruct vanished and abandoned(post)medieval routes and answer questions about human-landscape interactions.
CAA 2019 - 47th Annual Conference on Computer Applications and Quantitative Methods in Archaeology, 2019
Archaeologists have been adopting photogrammetry for photo-realistic 3D documentation for decades... more Archaeologists have been adopting photogrammetry for photo-realistic 3D documentation for decades by now. However, it is still up to the user to provide high quality images with good coverage and view overlap as input to achieve good results. This is not only time consuming but also requires a high degree of experience when capturing non-trivial archaeological features often found at excavation sites such as ditches or ceramics sherds with sharp edges. To overcome these limitations, I propose a novel approach derived from recent scientific progress in Robotics and Computer Vision. Instead of a human photographer, a computer decides autonomously which images should be taken while an object is captured. This is based on the actual shape of the object and promises to lead to substantially more accurate 3D results. To demonstrate the usefulness of the approach, I present two robot prototypes incorporating the underlying algorithms: one for entire sites (demonstrated with a Roman Villa Rustica) and one for smaller scale objects (here: Minoan pottery).
Proceedings of IEEE Symposium on Visual Languages, 1995
Programming-by-demonstration systems often have problems with control structure
inference and us... more Programming-by-demonstration systems often have problems with control structure
inference and user-intended generalization. We propose a new solution for these
weaknesses based on the concepts of the programming language AWK and present a
prototype system for text processing. It utilizes' vertical demonstration', extensive visual
feedback and program visualization via spreadsheets to achieve improved usability and
expressive power.
Proceedings of IEEE International Conference on Multimedia Computing and Systems, 1997
Virtual reality (VR) systems utilize more and more output media in order to make virtual environm... more Virtual reality (VR) systems utilize more and more output media in order to make virtual environments more realistic. Furthermore, VR developers nowadays concentrate on rich interaction and animation. This causes the need for a new VR development system architecture. Such a new architecture is proposed and discussed. Its key features include support for new output media, device independence, and rapid behavioral prototyping. The Lightning VR system is introduced as a prototype for the presented architecture. It has already shown its usefulness in several industrial applications.
Journal of Visual Languages & Computing, 1997
Programming by demonstration (PBD) promises to enable non-programmers to do more with their compu... more Programming by demonstration (PBD) promises to enable non-programmers to do more with their computers. However, PBD systems today are either limited to very narrow application domains, or are not reliable enough concerning the correctness of generated programs. We examine the reasons for this gap between generality and reliability and present a proposal for bridging it, thus leading to a design methodology for general-purpose PBD systems. This is achieved by supporting the visual scanning abilities of humans by a diligently designed user interface, immediate feedback, and ‘trial & error demonstration’. Furthermore, we discuss experiences with a prototype system named Visual AWK implementing the new concept. It enables end-users to create rather complex text-processing programs in a structure editor and thus automates repetitive tasks as they encounter them in their work.
Proceedings of the third symposium on Virtual reality modeling language, 1998
Uploads
books by Juergen Landauer
papers by Juergen Landauer
research is increasingly appreciated. In specific areas, notably remote sensing, datasets have increased
in extent and resolution to the point that manual interpretation is problematic and the availability
of skilled interpreters to undertake such work is limited. Interpretation of the geophysical datasets
associated with prehistoric submerged landscapes is particularly challenging. Following the Last
Glacial Maximum, sea levels rose by 120 m globally, and vast, habitable landscapes were lost to the
sea. These landscapes were inaccessible until extensive remote sensing datasets were provided by the
offshore energy sector. In this paper, we provide the results of a research programme centred on AI
applications using data from the southern North Sea. Here, an area of c. 188,000 km2 of habitable
terrestrial land was inundated between c. 20,000 BP and 7000 BP, along with the cultural heritage
it contained. As part of this project, machine learning tools were applied to detect and interpret
features with potential archaeological significance from shallow seismic data. The output provides a
proof-of-concept model demonstrating verifiable results and the potential for a further, more complex,
leveraging of AI interpretation for the study of submarine palaeolandscapes.
inference and user-intended generalization. We propose a new solution for these
weaknesses based on the concepts of the programming language AWK and present a
prototype system for text processing. It utilizes' vertical demonstration', extensive visual
feedback and program visualization via spreadsheets to achieve improved usability and
expressive power.
research is increasingly appreciated. In specific areas, notably remote sensing, datasets have increased
in extent and resolution to the point that manual interpretation is problematic and the availability
of skilled interpreters to undertake such work is limited. Interpretation of the geophysical datasets
associated with prehistoric submerged landscapes is particularly challenging. Following the Last
Glacial Maximum, sea levels rose by 120 m globally, and vast, habitable landscapes were lost to the
sea. These landscapes were inaccessible until extensive remote sensing datasets were provided by the
offshore energy sector. In this paper, we provide the results of a research programme centred on AI
applications using data from the southern North Sea. Here, an area of c. 188,000 km2 of habitable
terrestrial land was inundated between c. 20,000 BP and 7000 BP, along with the cultural heritage
it contained. As part of this project, machine learning tools were applied to detect and interpret
features with potential archaeological significance from shallow seismic data. The output provides a
proof-of-concept model demonstrating verifiable results and the potential for a further, more complex,
leveraging of AI interpretation for the study of submarine palaeolandscapes.
inference and user-intended generalization. We propose a new solution for these
weaknesses based on the concepts of the programming language AWK and present a
prototype system for text processing. It utilizes' vertical demonstration', extensive visual
feedback and program visualization via spreadsheets to achieve improved usability and
expressive power.
Photogrammetry has been adopted by archaeologists for documentation of artefacts for quite a while with good results. Even researchers with limited experience are able to produce good results by obeying simple rules such as “ensure sufficient overlap between images” or by using software tools such as Pix4D or Scann3D which incorporate such rules.
However, these rules are mere simplifications of the underlying photogrammetry algorithms. Whereas they work well with rather planar structures and objects without protruding or partially occluded features, they are bound to provide less-than-optimal results with a large set of real world archaeological objects.
Here, we propose a new approach derived from recent progress in robotics and computer vision (more specifically solutions to the so-called Next-Best-View problem) which lets the computer decide which image to take next based on previous images. By intertwining image capturing and processing a more accurate 3D documentation can be achieved.
We demonstrate that this method is useful for both entire sites (demonstrated with a Roman Villa Rustica) and smaller scale objects (e.g. Minoan pottery). Capturing the latter currently still requires some manual assistance, whereas site recordings are fully automatic (with drones).