Figure 1 | Scientific Reports

Figure 1

From: Author Correction: Ensembles of data-efficient vision transformers as a new paradigm for automated classification in ecology

Figure 1

Comparing EDeiTs to the previous SOTA. For each dataset, we show the error, which is the fraction of misclassified test images (\(1-accuracy\)). The error of the existing SOTA model is shown in orange. For the ensembles of DeiTs, we show two ways of combining the individual learnings: through arithmetic (blue) and geometric (purple) averaging. The purple bar for RSMAS is absent because all the test examples were classified correctly by the EDeit with geometric averaging. Independent of the ensembling rule, our models outperform current SOTA models on a consistent basis.

Back to article page