A5064838766- Species Distribution and Climate Change
- Plant and animal studies
- Advanced Vision and Imaging
- Advanced Image Processing Techniques
- Image Processing Techniques and Applications
- Plant Pathogens and Fungal Diseases
- Remote Sensing in Agriculture
- Genomics and Phylogenetic Studies
- Ecology and Vegetation Dynamics Studies
- Computer Graphics and Visualization Techniques
- Medical Image Segmentation Techniques
- Advanced Optical Sensing Technologies
- Remote Sensing and LiDAR Applications
- Opportunistic and Delay-Tolerant Networks
- AI in cancer detection
- Botany, Ecology, and Taxonomy Studies
- Identification and Quantification in Food
- 3D Shape Modeling and Analysis
- Generative Adversarial Networks and Image Synthesis
- Video Surveillance and Tracking Methods
- Plant Taxonomy and Phylogenetics
- Satellite Image Processing and Photogrammetry
Nvidia (United States)
2024
ETH Zurich
2021-2024
University of Zurich
2023
Abstract In the age of big data, scientific progress is fundamentally limited by our capacity to extract critical information. Here, we map fine-grained spatiotemporal distributions for thousands species, using deep neural networks (DNNs) and ubiquitous citizen science data. Based on 6.7 M observations, jointly model 2477 plant species aggregates across Switzerland with an ensemble DNNs built different cost functions. We find that, compared commonly-used approaches, multispecies predict...
Guided super-resolution is a unifying framework for several computer vision tasks where the inputs are low-resolution source image of some target quantity (e.g., perspective depth acquired with time-of-flight camera) and high-resolution guide from different domain grey-scale conventional camera); output version (in our example, high-res map). The standard way looking at this problem to formulate it as task, i.e., upsampled resolution, while transferring missing high-frequency details guide....
We introduce a novel formulation for guided super-resolution. Its core is differentiable optimisation layer that operates on learned affinity graph. The graph potentials make it possible to leverage rich contextual information from the guide image, while explicit within architecture guarantees rigorous fidelity of high-resolution target low-resolution source. With decision employ source as constraint rather than only an input prediction, our method differs state-of-the-art deep architectures...
Particle-based representations of radiance fields such as 3D Gaussian Splatting have found great success for reconstructing and re-rendering complex scenes. Most existing methods render particles via rasterization, projecting them to screen space tiles processing in a sorted order. This work instead considers ray tracing the particles, building bounding volume hierarchy casting each pixel using high-performance GPU hardware. To efficiently handle large numbers semi-transparent we describe...
Automatic identification of plant specimens from amateur photographs could improve species range maps, thus supporting ecosystems research as well conservation efforts.However, classifying based on image data alone is challenging: some exhibit large variations in visual appearance, while at the same time different are often visually similar; additionally, observations follow a highly imbalanced, long-tailed distribution due to differences abundance observer biases.On other hand, most...
Herbarium sheets present a unique view of the world's botanical history, evolution, and biodiversity. This makes them an all-important data source for research. With increased digitization herbaria worldwide advances in domain fine-grained visual classification which can facilitate automatic identification herbarium specimen images, there are many opportunities supporting expanding research this field. However, existing datasets either too small, or not diverse enough, terms represented...
We introduce OmniRe, a holistic approach for efficiently reconstructing high-fidelity dynamic urban scenes from on-device logs. Recent methods modeling driving sequences using neural radiance fields or Gaussian Splatting have demonstrated the potential of challenging scenes, but often overlook pedestrians and other non-vehicle actors, hindering complete pipeline scene reconstruction. To that end, we propose comprehensive 3DGS framework named allows accurate, full-length reconstruction...
<title>Abstract</title> In the age of big data, scientific progress is fundamentally limited by our capacity to extract critical information. We show that recasting multispecies distribution modeling as a ranking problem allows analyzing ubiquitous citizen-science observations with unprecedented efficiency. Based on 6.7M observations, we jointly modeled distributions 2477 plant species and aggregates across Switzerland, using deep neural networks (DNNs). Compared commonly-used approaches,...
Neural reconstruction approaches are rapidly emerging as the preferred representation for 3D scenes, but their limited editability is still posing a challenge. In this work, we propose an approach scene inpainting -- task of coherently replacing parts reconstructed with desired content. Scene inherently ill-posed there exist many solutions that plausibly replace missing A good method should therefore not only enable high-quality synthesis also high degree control. Based on observation, focus...
Citizen science has become key to biodiversity monitoring but critically depends on accurate quality control that is scalable and tailored the focal region. We developed FlorID, a free-to-use identification service for all native many non-native plants of Switzerland. FlorID can identify >3000 species, using vision transformers trained 1.5M photos, ecological predictions from multilayer perceptrons, 6.7M occurrence observations 20 high-resolution environmental variables. Embedded in...
Particle-based representations of radiance fields such as 3D Gaussian Splatting have found great success for reconstructing and re-rendering complex scenes. Most existing methods render particles via rasterization, projecting them to screen space tiles processing in a sorted order. This work instead considers ray tracing the particles, building bounding volume hierarchy casting each pixel using high-performance GPU hardware. To efficiently handle large numbers semi-transparent we describe...
Abstract In the age of big data, scientific progress is fundamentally limited by our capacity to extract critical information. We show that recasting multispecies distribution modeling as a ranking problem allows analyzing ubiquitous citizen-science observations with unprecedented efficiency. Based on 6.7M observations, we jointly modeled distributions 2477 plant species and aggregates across Switzerland, using deep neural networks (DNNs). Compared commonly-used approaches, DNNs predicted...
Herbarium sheets present a unique view of the world's botanical history, evolution, and diversity. This makes them an all-important data source for research. With increased digitisation herbaria worldwide advances in fine-grained classification domain that can facilitate automatic identification herbarium specimens, there are lot opportunities supporting research this field. However, existing datasets either too small, or not diverse enough, terms represented taxa, geographic distribution...
Guided super-resolution is a unifying framework for several computer vision tasks where the inputs are low-resolution source image of some target quantity (e.g., perspective depth acquired with time-of-flight camera) and high-resolution guide from different domain grey-scale conventional camera); output version (in our example, high-res map). The standard way looking at this problem to formulate it as task, i.e., upsampled resolution, while transferring missing high-frequency details guide....
We introduce a novel formulation for guided super-resolution. Its core is differentiable optimisation layer that operates on learned affinity graph. The graph potentials make it possible to leverage rich contextual information from the guide image, while explicit within architecture guarantees rigorous fidelity of high-resolution target low-resolution source. With decision employ source as constraint rather than only an input prediction, our method differs state-of-the-art deep architectures...
Automatic identification of plant specimens from amateur photographs could improve species range maps, thus supporting ecosystems research as well conservation efforts. However, classifying based on image data alone is challenging: some exhibit large variations in visual appearance, while at the same time different are often visually similar; additionally, observations follow a highly imbalanced, long-tailed distribution due to differences abundance observer biases. On other hand, most...