Multi-Scale ViT Inference with Habitat-Fit Priors and kNN Retrieval for Multi-Species Plant Identification

2026-07-15

AI Summary

This study investigates how models trained exclusively on single-species images can effectively identify all co-occurring plant species in high-resolution vegetation plot imagery. The approach builds upon a fine-tuned DINOv2 ViT-L/14 backbone and integrates multi-scale patch-based inference, FAISS kNN retrieval, habitat-adaptive reweighting informed by geographic and elevational priors, temporal-aware prediction fusion, and region masking in post-processing. The core contributions lie in the proposed habitat-adaptive reweighting mechanism and a multi-scale aggregation strategy that synergistically combine external ecological priors with visual features. Evaluated on the PlantCLEF 2026 private test set, the method achieves a macro-F1 score of 0.43902, ranking third overall, while an unsubmitted configuration exceeds 0.45.

Abstract

This paper describes DS@GT ARC's third-place solution to the PlantCLEF 2026 challenge on multi-species plant identification in vegetation quadrat images, where systems must predict every species present in high-resolution (~3000 x 3000 pixel) plot photographs while training only on single-label images of individual plants. The pipeline is built around a fine-tuned DINOv2 ViT-L/14 classifier applied over a multi-scale tile decomposition of each quadrat, with per-tile predictions blended with a FAISS kNN retriever and post-processed by source-aware temporal fusion across repeated plot visits, a habitat-fit demotion that injects geographic and altitude priors from the training data, and a South-Western Europe geographic mask. Habitat-fit demotion and multi-scale aggregation are the largest individual contributors in the ablations. Two complementary training-centric directions, a cross-region transformer with noisy-student distillation on the LUCAS dataset and a label-as-query transformer decoder over synthetic CLS-domain pseudo-quadrats, yielded null results. An inference-time augmentation with instance-aware segmentation crops also did not improve performance. The selected submission reaches a private-leaderboard macro-F1 of 0.43902 (third place; public 0.51096); an unselected configuration of the same pipeline scored above 0.45 on the private set. Code: https://github.com/dsgt-arc/plantclef-2026.

Problem

• multi-species plant identification
• vegetation quadrat images
• single-label training
• high-resolution image analysis
• PlantCLEF

Innovation

• Multi-Scale ViT
• Habitat-Fit Priors
• kNN Retrieval
• Quadrat-Based Plant Identification
• DINOv2 Fine-Tuning