Back to Blog
Dense open set contains a subshift5/5/2023 Previous work has addressed dense anomaly detection by. This problem is especially demanding in the context of dense prediction since input images may be partially anomalous. The resulting models often give rise to confident incorrect predictions which may lead to devastating consequences. Standard machine learning is unable to accommodate inputs which do not belong to the training distribution. Official implementation: anomaly detectionDense open-set recognitionOut-of-distribution detectionSemantic segmentation Read more Our submissions achieve state-of-the-art performance despite neglectable computational overhead over the standard semantic segmentation baseline. Experiments evaluate our contributions on standard dense anomaly detection benchmarks as well as in terms of open-mIoU - a novel metric for dense open-set performance. We carefully design a custom loss for the data likelihood in order to avoid backpropagation through the untractable normalizing constant \(Z(\theta )\). We blend these two predictions into a hybrid anomaly score which allows dense open-set recognition on large natural images. The latter two predictions are trained both on the standard training data and on a generic negative dataset. We therefore design a novel hybrid algorithm based on reinterpreting discriminative logits as a logarithm of the unnormalized joint distribution $\hat)\). These requirements disqualify all previous hybrid approaches to the best of our knowledge. Unfortunately, dense anomaly detection requires translational equivariance and very large input resolutions. Consequently, hybrid algorithms present an attractive research goal. These two approaches exhibit different failure modes. Anomaly detection can be conceived either through generative modelling of regular training data or by discriminating with respect to negative training data.
0 Comments
Read More
Leave a Reply. |