We present HUDD, a tool that supports safety analysis practices for systems integrating Deep Neural Networks (DNNs) by automatically identifying the root causes for DNN errors and retraining the DNN. Its name stands for Heatmap-based Unsupervised Debugging of DNNs, it automatically clusters together error-inducing images whose results are due to common subsets of DNN neurons. The generated clusters group error-inducing images having common characteristics, that is, having a common root cause.

HUDD identifies root causes by applying a clustering algorithm to matrices (i.e., heatmaps) capturing the relevance of every DNN neuron on the DNN outcome. Also, HUDD retrains DNNs with images that are automatically selected based on their relatedness to the identified image clusters. Our empirical evaluation with DNNs from the automotive domain have shown that HUDD automatically identifies all the distinct root causes of DNN errors, thus supporting safety analysis. Also, our retraining approach has shown to be more effective at improving DNN accuracy than existing approaches.

A demo video of HUDD is available at https://youtu.be/drjVakP7jdU