Does SAM dream of EIG? Characterizing Interactive Segmenter Performance using Expected Information Gain
This work addresses the need for better evaluation metrics in interactive segmentation, which is important for researchers and practitioners in computer vision, though it is incremental as it builds on existing Bayesian concepts.
The authors tackled the problem of assessing interactive segmentation models by introducing a procedure based on Bayesian Experimental Design to measure how well models understand point prompts and their correspondence with segmentation masks, showing that Oracle Dice index measurements are insensitive or misleading for this property.
We introduce an assessment procedure for interactive segmentation models. Based on concepts from Bayesian Experimental Design, the procedure measures a model's understanding of point prompts and their correspondence with the desired segmentation mask. We show that Oracle Dice index measurements are insensitive or even misleading in measuring this property. We demonstrate the use of the proposed procedure on three interactive segmentation models and subsets of two large image segmentation datasets.