African or European Swallow? Benchmarking Large Vision-Language Models for Fine-Grained Object Classification

TL;DR

This paper introduces FOCI, a challenging benchmark for fine-grained object classification, revealing that current LVLMs underperform compared to CLIP models, highlighting the need for better fine-grained alignment in training.

Contribution

The paper creates a new benchmark, FOCI, for evaluating LVLMs on fine-grained classification, and demonstrates the performance gap with CLIP models, emphasizing the need for improved fine-grained training data.

Findings

CLIP models outperform LVLMs significantly on FOCI.

FOCI tests for a complementary skill to existing benchmarks.

LVLMs' image encoders lack fine-grained alignment.

Abstract

Recent Large Vision-Language Models (LVLMs) demonstrate impressive abilities on numerous image understanding and reasoning tasks. The task of fine-grained object classification (e.g., distinction between \textit{animal species}), however, has been probed insufficiently, despite its downstream importance. We fill this evaluation gap by creating \texttt{FOCI} (\textbf{F}ine-grained \textbf{O}bject \textbf{C}lass\textbf{I}fication), a difficult multiple-choice benchmark for fine-grained object classification, from existing object classification datasets: (1) multiple-choice avoids ambiguous answers associated with casting classification as open-ended QA task; (2) we retain classification difficulty by mining negative labels with a CLIP model. \texttt{FOCI}\xspace complements five popular classification datasets with four domain-specific subsets from ImageNet-21k. We benchmark 12 public LVLMs on \texttt{FOCI} and show that it tests for a \textit{complementary skill} to established image understanding and reasoning benchmarks. Crucially, CLIP models exhibit dramatically better performance than LVLMs. Since the image encoders of LVLMs come from these CLIP models, this points to inadequate alignment for fine-grained object distinction between the encoder and the LLM and warrants (pre)training data with more fine-grained annotation. We release our code at \url{https://github.com/gregor-ge/FOCI-Benchmark}.