Search Results for author: Steve Dias Da Cruz

Found 7 papers, 3 papers with code

Autoencoder for Synthetic to Real Generalization: From Simple to More Complex Scenes

1 code implementation1 Apr 2022 Steve Dias Da Cruz, Bertram Taetz, Thomas Stifter, Didier Stricker

Learning on synthetic data and transferring the resulting properties to their real counterparts is an important challenge for reducing costs and increasing safety in machine learning.

Autoencoder Attractors for Uncertainty Estimation

1 code implementation1 Apr 2022 Steve Dias Da Cruz, Bertram Taetz, Thomas Stifter, Didier Stricker

While input images close to known samples will converge to the same or similar attractor, input samples containing unknown features are unstable and converge to different training samples by potentially removing or changing characteristic features.

Gaussian Processes

Topological properties of basins of attraction and expressiveness of width bounded neural networks

no code implementations10 Nov 2020 Hans-Peter Beise, Steve Dias Da Cruz

In Radhakrishnan et al. [2020], the authors empirically show that autoencoders trained with usual SGD methods shape out basins of attraction around their training data.

Illumination Normalization by Partially Impossible Encoder-Decoder Cost Function

no code implementations6 Nov 2020 Steve Dias Da Cruz, Bertram Taetz, Thomas Stifter, Didier Stricker

Our method exploits the availability of identical sceneries under different illumination and environmental conditions for which we formulate a partially impossible reconstruction target: the input image will not convey enough information to reconstruct the target in its entirety.

SVIRO: Synthetic Vehicle Interior Rear Seat Occupancy Dataset and Benchmark

1 code implementation10 Jan 2020 Steve Dias Da Cruz, Oliver Wasenmüller, Hans-Peter Beise, Thomas Stifter, Didier Stricker

We release SVIRO, a synthetic dataset for sceneries in the passenger compartment of ten different vehicles, in order to analyze machine learning-based approaches for their generalization capacities and reliability when trained on a limited number of variations (e. g. identical backgrounds and textures, few instances per class).

Instance Segmentation object-detection +3

On decision regions of narrow deep neural networks

no code implementations3 Jul 2018 Hans-Peter Beise, Steve Dias Da Cruz, Udo Schröder

We show that for neural network functions that have width less or equal to the input dimension all connected components of decision regions are unbounded.

Cannot find the paper you are looking for? You can Submit a new open access paper.