Tile-Based ViT Inference with Visual-Cluster Priors for Zero-Shot Multi-Species Plant Identification

TL;DR

This paper presents a novel tile-based Vision Transformer approach with visual-cluster priors for zero-shot multi-species plant identification, achieving competitive results without additional training.

Contribution

It introduces a tiling strategy combined with visual clustering and Bayesian priors for improved inference in plant identification tasks.

Findings

Achieved macro-averaged F1 of 0.348 on private leaderboard

Utilized a 4x4 tiling strategy aligned with network receptive field

No additional training required for the proposed method

Abstract

We describe DS@GT's second-place solution to the PlantCLEF 2025 challenge on multi-species plant identification in vegetation quadrat images. Our pipeline combines (i) a fine-tuned Vision Transformer ViTD2PC24All for patch-level inference, (ii) a 4x4 tiling strategy that aligns patch size with the network's 518x518 receptive field, and (iii) domain-prior adaptation through PaCMAP + K-Means visual clustering and geolocation filtering. Tile predictions are aggregated by majority vote and re-weighted with cluster-specific Bayesian priors, yielding a macro-averaged F1 of 0.348 (private leaderboard) while requiring no additional training. All code, configuration files, and reproducibility scripts are publicly available at https://github.com/dsgt-arc/plantclef-2025.