AI model reconstructs missing parts of leaf fossils

Researchers have developed AmodalDINO, a novel multi-head dense-prediction model designed for the amodal reconstruction of leaf fossil images. This model can predict the visible leaf, the complete amodal leaf, the main vein, and fine veins from a single RGB image without requiring an upstream instance segmenter. By fine-tuning a DINOv3 ViT-L/16 model and adding auxiliary venation heads, AmodalDINO learns the structural shape prior of leaves, achieving high accuracy on synthetic and real fossil specimens. The model is also efficient, capable of running offline in a browser after quantization to 4-bit weights, and includes features for estimating surface area and visualizing living leaves. AI

IMPACT This research demonstrates a novel application of AI for paleontology, potentially improving the analysis of fossilized remains.

RANK_REASON The cluster contains an academic paper detailing a new AI model and its performance on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• AmodalDINO
• DINOv3 ViT-L/16

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →