Report on the maturation of local characterisation tools
Owner
Description
The contributions in this deliverable are divided into 4 different parts. The first three correspond to the three tools matured in this action sheet, while the last one correspond to the evaluation of Chiru. A quick summary of the different tool and use cases tested is given below:
Chapter 2: AI Metamorphism Observing Software from CEA
Tested on: UC Renault Welding, UC Safran Visual Industrial Control
Short description: Assesses metamorphic properties on AI models such as robustness to perturbations on the inputs but also relation between models’ inputs and outputs.
Chapter 3: Amplification methods for robustness evaluation from LNE
Tested on: UC Valeo Scene understanding, UC Air Liquid Demand Forecasting, UC Safran Visual Industrial Control
Short description: Evaluates the robustness of models using amplification methods on the dataset with noise functions.
Chapter 4: Non-overlapping corruption benchmarking tool from Atos
Tested on: UC Atos ReID
Short description: Provides a benchmark of synthetic corruptions on a dataset to assess the robustness of a given AI-based model.
Files
Files
Restricted
The record is publicly accessible, but files are restricted to users with access.