LASA: New Weak Supervision Method for Open-Vocabulary Sketch Semantic Segmentation
Researchers introduced LASA, a method for assigning semantic labels to sketch drawings without pixel-level training annotations, using multi-layer Vision Transformer attention maps. The approach aggregates attention from different transformer layers to capture both global structure and local details, addressing the challenge that sketches lack texture and color cues. This work advances open-vocabulary segmentation for sparse line drawings, which has applications in sketch-based image retrieval and design automation.
LASA (Layer-wise Accumulated Structural Attention) is a weak supervision framework designed for open-vocabulary scene sketch semantic segmentation. The method addresses a key limitation of existing approaches: single-layer vision-language features are unstable for sketch understanding because sketches depend heavily on stroke layout and spatial configuration rather than texture or color. The researchers observed that different Vision Transformer layers encode complementary information—shallow layers capture global structural layouts while deeper layers focus on local stroke intersections and object parts. By aggregating multi-layer attention maps, LASA provides a more robust structural prior. Experiments on three benchmarks (FS-COCO, SFSD, and FrISS) demonstrated substantial improvements, with mIoU gains of +3.43, +8.01, and +15.74 points respectively over prior weakly supervised baselines. The authors committed to releasing source code publicly.
What different sources said
- arXiv cs.AICenter
LASA: A Weak Supervision Method for Open-Vocabulary Scene Sketch Semantic Segmentation
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.