Add to ORKGRecurrent models are a popular choice for video enhancement tasks such as video denoising or super-resolution. In this work, we focus on their stability as dynamical systems and show that they tend to fail catastrophically at inference time on long video sequences. To address this issue, we (1) introduce a diagnostic tool which produces input sequences optimized to trigger instabilities and that can be interpreted as visualizations of temporal receptive fields, and (2) propose two approaches to enforce the stability of a model during training: constraining the spectral norm or constraining the stable rank of its convolutional layers. We then introduce Stable Rank Normalization for Convolutional layers (SRN-C), a new algorithm that enforces ...
In this paper we address the issue of output instability of deep neural networks: small perturbation...
Traditionally, convolutional neural networks are feedforward networks with a deep and complex hierar...
Applying convolutional neural networks to large images is computationally ex-pensive because the amo...
Recurrent models are a popular choice for video enhancement tasks such as video denoising or super-r...
This dissertation provides a generic solution to model dynamic systems whose hidden state and the tr...
Convolutional neural networks (CNNs) can model complicated non-linear relations between images. Howe...
International audienceWhen trying to independently apply image-trained algorithms to successive fram...
International audienceExtending image processing techniques to videos is a non-trivial task; applyin...
This paper studies the dynamic generator model for spatialtemporal processes such as dynamic texture...
In this chapter, we describe the basic concepts behind the functioning of recurrent neural networks ...
Figure 1: With many image filters, such as this automatic color, tone, and contrast adjustment, proc...
Video image recognition has been extensively studied with rapid progress recently. However, most met...
The paper first summarizes a general approach to the training of recurrent neural networks by gradie...
This work introduces double-mapping Gated Recurrent Units (dGRU), an extension of standard GRUs wher...
The temporal events in video sequences often have long-term dependencies which are difficult to be h...
In this paper we address the issue of output instability of deep neural networks: small perturbation...
Traditionally, convolutional neural networks are feedforward networks with a deep and complex hierar...
Applying convolutional neural networks to large images is computationally ex-pensive because the amo...
Recurrent models are a popular choice for video enhancement tasks such as video denoising or super-r...
This dissertation provides a generic solution to model dynamic systems whose hidden state and the tr...
Convolutional neural networks (CNNs) can model complicated non-linear relations between images. Howe...
International audienceWhen trying to independently apply image-trained algorithms to successive fram...
International audienceExtending image processing techniques to videos is a non-trivial task; applyin...
This paper studies the dynamic generator model for spatialtemporal processes such as dynamic texture...
In this chapter, we describe the basic concepts behind the functioning of recurrent neural networks ...
Figure 1: With many image filters, such as this automatic color, tone, and contrast adjustment, proc...
Video image recognition has been extensively studied with rapid progress recently. However, most met...
The paper first summarizes a general approach to the training of recurrent neural networks by gradie...
This work introduces double-mapping Gated Recurrent Units (dGRU), an extension of standard GRUs wher...
The temporal events in video sequences often have long-term dependencies which are difficult to be h...
In this paper we address the issue of output instability of deep neural networks: small perturbation...
Traditionally, convolutional neural networks are feedforward networks with a deep and complex hierar...
Applying convolutional neural networks to large images is computationally ex-pensive because the amo...