With a sensitive statistical test, such as KS(conf), it is possible to detect even subtle changes to the input images from the distribution of output scores. This includes cases that are hardly visible to the human eye, such as weak Gaussian smoothing, a small number of dead pixels, or a small amount of additive noise.
Feb 14, 2019
Apr 11, 2018 · In this work, we describe KS(conf), a procedure for detecting such outside of the specifications operation. Building on statistical insights, ...
Automatic test if a classifier operates outside of its specifications: ○ justified statistically. ○ easy to implement. ○ applicable for any classifier that ...
In this work, we describe KS(conf), a procedure for detecting such outside of the specifications operation. Building on statistical insights, its main step is ...
KS(conf): A Light-Weight Test if a ConvNet Operates Outside of Its Specifications. https://doi.org/10.1007/978-3-030-12939-2_18 · Full text.
We study the problem of automatically detecting if a given multi-class classifier operates outside of its specifications (out-of-specs).
In this work, we describe KS(conf), a procedure for detecting such outside of specifications (out-of-specs) operation, based on statistical testing of the ...
KS(conf): A light-weight test if a ConvNet operates outside of Its specifications · Author. Sun, Rémy; Lampert , Christoph · Department. Lampert Group · Grant.
A Light-Weight Test if a ConvNet Operates Outside of Its Specifications. Code and example data for KS(conf) test and baselines. For examples of use, ...
Oct 10, 2019 · We study the problem of automatically detecting if a given multi-class classifier operates outside of its specifications (out-of-specs).
Missing: ConvNet | Show results with:ConvNet