On the Minimal Recognizable Image Patch
This addresses a limitation in computer vision for recognition tasks, but it is incremental as it builds on existing human vision research.
The paper tackles the problem of algorithmic failure on partially occluded images by empirically characterizing the minimal recognizable patch (MRP) sufficient for recognition, finding that sharp reductions in accuracy occur with size reduction, similar to human vision studies.
In contrast to human vision, common recognition algorithms often fail on partially occluded images. We propose characterizing, empirically, the algorithmic limits by finding a minimal recognizable patch (MRP) that is by itself sufficient to recognize the image. A specialized deep network allows us to find the most informative patches of a given size, and serves as an experimental tool. A human vision study recently characterized related (but different) minimally recognizable configurations (MIRCs) [1], for which we specify computational analogues (denoted cMIRCs). The drop in human decision accuracy associated with size reduction of these MIRCs is substantial and sharp. Interestingly, such sharp reductions were also found for the computational versions we specified.