Selected work

Representative Papers

Atomizer-IO: Beyond Pixels, Patches and Grids

Sep 30, 2026

This study addresses the limitation of existing vision architectures that rely on regular grid assumptions, rendering them ineffective for unstructured sensor data. To overcome this, we propose a novel architecture based on atomic representations, introducing an "observation-first" paradigm. By leveraging measurement metadata and anchor-based local cross-attention mechanisms, our approach infers structural information directly from physical relationships, thereby eliminating grid constraints entirely. The proposed architecture seamlessly generalizes to diverse inputs, such as unordered 3D point clouds, without requiring task-specific redesigns. Furthermore, it achieves performance levels comparable to specialized models across multiple tasks while enabling post-training control over inference costs. These results comprehensively validate the versatility and flexibility of the proposed grid-free interface for processing heterogeneous sensory data.

0 citationsRead paper

Overview of LifeCLEF Plant Identification task 2020

Sep 23, 2025

Automatic plant identification in tropical regions—exemplified by the Guiana Shield in South America, hosting ~1,000 highly diverse plant species—is severely constrained by the scarcity of field-collected image data. Method: This study systematically evaluates, for the first time, the feasibility of leveraging digitized herbarium specimens to enhance cross-domain recognition performance. We formulate a large-scale herbarium-to-field-image classification task and employ deep convolutional neural networks to learn feature mappings between herbarium sheets and in-situ photographs, enabling effective knowledge transfer. Contribution/Results: By integrating state-of-the-art models from multiple research teams, we demonstrate that combining a small number of field images with abundant herbarium data significantly improves classification accuracy on real-world photographs. This work bridges classical botanical resources with modern deep learning, establishing a scalable, transferable technical paradigm for intelligent biodiversity monitoring in data-scarce regions.

0 citationsRead paper

Recent publications

Latest Papers

Atomizer-IO: Beyond Pixels, Patches and Grids

Sep 30, 2026

This study addresses the limitation of existing vision architectures that rely on regular grid assumptions, rendering them ineffective for unstructured sensor data. To overcome this, we propose a novel architecture based on atomic representations, introducing an "observation-first" paradigm. By leveraging measurement metadata and anchor-based local cross-attention mechanisms, our approach infers structural information directly from physical relationships, thereby eliminating grid constraints entirely. The proposed architecture seamlessly generalizes to diverse inputs, such as unordered 3D point clouds, without requiring task-specific redesigns. Furthermore, it achieves performance levels comparable to specialized models across multiple tasks while enabling post-training control over inference costs. These results comprehensively validate the versatility and flexibility of the proposed grid-free interface for processing heterogeneous sensory data.

0 citationsRead paper

Overview of LifeCLEF Plant Identification task 2020

Sep 23, 2025

Automatic plant identification in tropical regions—exemplified by the Guiana Shield in South America, hosting ~1,000 highly diverse plant species—is severely constrained by the scarcity of field-collected image data. Method: This study systematically evaluates, for the first time, the feasibility of leveraging digitized herbarium specimens to enhance cross-domain recognition performance. We formulate a large-scale herbarium-to-field-image classification task and employ deep convolutional neural networks to learn feature mappings between herbarium sheets and in-situ photographs, enabling effective knowledge transfer. Contribution/Results: By integrating state-of-the-art models from multiple research teams, we demonstrate that combining a small number of field images with abundant herbarium data significantly improves classification accuracy on real-world photographs. This work bridges classical botanical resources with modern deep learning, establishing a scalable, transferable technical paradigm for intelligent biodiversity monitoring in data-scarce regions.

0 citationsRead paper