Structuring Autoencoders

Rudolph, Marco; Wandt, Bastian; Rosenhahn, Bodo

doi:10.1109/iccvw.2019.00075

Cited by 29 publications

(10 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The method itself is not bound to faces or images at all. We also want to make NNs more interpretable and show the importance of such interpretations by working on both the forced disentanglement presented in this paper and also unsupervised disentanglement as shown with Structuring Autoencoders [28]. We are also interested in making our method non-deterministic, similar to a Markov Chain Neural Network [1] and we want to adapt our method to other data sets in 3D [33].…”

Section: Discussionmentioning

confidence: 99%

Learning Disentangled Representations via Independent Subspaces

Awiszus

Ackermann

Rosenhahn

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

Self Cite

View full text Add to dashboard Cite

Image generating neural networks are mostly viewed as black boxes, where any change in the input can have a number of globally effective changes on the output. In this work, we propose a method for learning disentangled representations to allow for localized image manipulations. We use face images as our example of choice. Depending on the image region, identity and other facial attributes can be modified. The proposed network can transfer parts of a face such as shape and color of eyes, hair, mouth, etc. directly between persons while all other parts of the face remain unchanged. The network allows to generate modified images which appear like realistic images. Our model learns disentangled representations by weak supervision. We propose a localized resnet autoencoder optimized using several loss functions including a loss based on the semantic segmentation, which we interpret as masks, and a loss which enforces disentanglement by decomposition of the latent space into statistically independent subspaces. We evaluate the proposed solution w.r.t. disentanglement and generated image quality. Convincing results are demonstrated using the CelebA dataset [24].

show abstract

Section: Discussionmentioning

confidence: 99%

Learning Disentangled Representations via Independent Subspaces

Awiszus

Ackermann

Rosenhahn

2019

2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Many anomaly detection methods are based on generative models, such as autoencoders [24,21,30] and GANs [16], which are optimized to generate the normal data. These approaches detect anomalies by the inability of the generative model to reconstruct them.…”

Section: Generative Modelsmentioning

confidence: 99%

Fully Convolutional Cross-Scale-Flows for Image-based Defect Detection

Rudolph¹,

Wehrbein²,

Rosenhahn³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

In industrial manufacturing processes, errors frequently occur at unpredictable times and in unknown manifestations. We tackle the problem of automatic defect detection without requiring any image samples of defective parts. Recent works model the distribution of defect-free image data, using either strong statistical priors or overly simplified data representations. In contrast, our approach handles fine-grained representations incorporating the global and local image context while flexibly estimating the density. To this end, we propose a novel fully convolutional cross-scale normalizing flow (CS-Flow) that jointly processes multiple feature maps of different scales. Using normalizing flows to assign meaningful likelihoods to input samples allows for efficient defect detection on image-level. Moreover, due to the preserved spatial arrangement the latent space of the normalizing flow is interpretable which enables to localize defective regions in the image. Our work sets a new stateof-the-art in image-level defect detection on the benchmark datasets Magnetic Tile Defects and MVTec AD showing a 100% AUROC on 4 out of 15 classes.

show abstract

“…Modern variants of AE are able to enforce some distribution in latent space either using the variational inference approach [5,11] or using the auxiliary discriminator network [9]. The work [12] is most close to ours in the sight of autoencoder based representation learning. The authors used an approach which is capable of building informative representations and visualization of the data.…”

Section: Related Workmentioning

confidence: 99%

Continuous Histogram Loss: Beyond Neural Similarity

Zholus¹,

Putin²

2020

Preprint

View full text Add to dashboard Cite

Similarity learning has gained a lot of attention from researches in recent years and tons of successful approaches have been recently proposed. However, the majority of the state-of-the-art similarity learning methods consider only a binary similarity. In this paper we introduce a new loss function called Continuous Histogram Loss (CHL) which generalizes recently proposed Histogram loss to multiple-valued similarities, i.e. allowing the acceptable values of similarity to be continuously distributed within some range. The novel loss function is computed by aggregating pairwise distances and similarities into 2D histograms in a differentiable manner and then computing the probability of condition that pairwise distances will not decrease as the similarities increase. The novel loss is capable of solving a wider range of tasks including similarity learning, representation learning and data visualization.

show abstract

Structuring Autoencoders

Cited by 29 publications

References 19 publications

Learning Disentangled Representations via Independent Subspaces

Learning Disentangled Representations via Independent Subspaces

Fully Convolutional Cross-Scale-Flows for Image-based Defect Detection

Continuous Histogram Loss: Beyond Neural Similarity

Contact Info

Product

Resources

About