arXiv subjects

Radu Timofte

Publications and source records attributed to Radu Timofte.

At least 361 records · Page 20Linked to original sources

arXiv2019

Learning Filter Basis for Convolutional Neural Network Compression

Convolutional neural networks (CNNs) based solutions have achieved state-of-the-art performances for many computer vision tasks, including classification and super-resolution of images. Usually the success of these methods comes with a cost of millions of parameters due to stacking deep convolutional layers. Moreover, quite a large number of filters are also used for a single convolutional layer, which exaggerates the parameter burden of current methods. Thus, in this paper, we try to reduce the number of parameters of CNNs by learning a basis of the filters in convolutional layers. For the forward pass, the learned basis is used to approximate the original filters and then used as parameters for the convolutional layers. We validate our proposed solution for multiple CNN architectures on image classification and image super-resolution benchmarks and compare favorably to the existing state-of-the-art in terms of reduction of parameters and preservation of accuracy.

cs.CV↗

arXiv2019

Adversarial Feature Distribution Alignment for Semi-Supervised Learning

Training deep neural networks with only a few labeled samples can lead to overfitting. This is problematic in semi-supervised learning where only a few labeled samples are available. In this paper, we show that a consequence of overfitting in SSL is feature distribution misalignment between labeled and unlabeled samples. Hence, we propose a new feature distribution alignment method. Our method is particularly effective when using only a small amount of labeled samples. We test our method on CIFAR10 and SVHN. On SVHN we achieve a test error of 3.88% (250 labeled samples) and 3.39% (1000 labeled samples) which is close to the fully supervised model 2.89% (73k labeled samples). In comparison, the current SOTA achieves only 4.29% and 3.74%. Finally, we provide a theoretical insight why feature distribution alignment occurs and show that our method reduces it.

cs.CV↗

arXiv2019

Adversarial Sampling for Active Learning

This paper proposes asal, a new GAN based active learning method that generates high entropy samples. Instead of directly annotating the synthetic samples, ASAL searches similar samples from the pool and includes them for training. Hence, the quality of new samples is high and annotations are reliable. To the best of our knowledge, ASAL is the first GAN based AL method applicable to multi-class problems that outperforms random sample selection. Another benefit of ASAL is its small run-time complexity (sub-linear) compared to traditional uncertainty sampling (linear). We present a comprehensive set of experiments on multiple traditional data sets and show that ASAL outperforms similar methods and clearly exceeds the established baseline (random sampling). In the discussion section we analyze in which situations ASAL performs best and why it is sometimes hard to outperform random sample selection.

cs.LG↗

arXiv2019

AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Results

This paper reviews the AIM 2019 challenge on real world super-resolution. It focuses on the participating methods and final results. The challenge addresses the real world setting, where paired true high and low-resolution images are unavailable. For training, only one set of source input images is therefore provided in the challenge. In Track 1: Source Domain the aim is to super-resolve such images while preserving the low level image characteristics of the source input domain. In Track 2: Target Domain a set of high-quality images is also provided for training, that defines the output domain and desired quality of the super-resolved images. To allow for quantitative evaluation, the source input images in both tracks are constructed using artificial, but realistic, image degradations. The challenge is the first of its kind, aiming to advance the state-of-the-art and provide a standard benchmark for this newly emerging task. In total 7 teams competed in the final testing phase, demonstrating new and innovative solutions to the problem.

cs.CV↗

arXiv2019

Frequency Separation for Real-World Super-Resolution

Most of the recent literature on image super-resolution (SR) assumes the availability of training data in the form of paired low resolution (LR) and high resolution (HR) images or the knowledge of the downgrading operator (usually bicubic downscaling). While the proposed methods perform well on standard benchmarks, they often fail to produce convincing results in real-world settings. This is because real-world images can be subject to corruptions such as sensor noise, which are severely altered by bicubic downscaling. Therefore, the models never see a real-world image during training, which limits their generalization capabilities. Moreover, it is cumbersome to collect paired LR and HR images in the same source domain. To address this problem, we propose DSGAN to introduce natural image characteristics in bicubically downscaled images. It can be trained in an unsupervised fashion on HR images, thereby generating LR images with the same characteristics as the original images. We then use the generated data to train a SR model, which greatly improves its performance on real-world images. Furthermore, we propose to separate the low and high image frequencies and treat them differently during training. Since the low frequencies are preserved by downsampling operations, we only require adversarial training to modify the high frequencies. This idea is applied to our DSGAN model as well as the SR model. We demonstrate the effectiveness of our method in several experiments through quantitative and qualitative analysis. Our solution is the winner of the AIM Challenge on Real World SR at ICCV 2019.

eess.IV↗

arXiv2019

AIM 2019 Challenge on Image Demoireing: Methods and Results

This paper reviews the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019. This paper describes the challenge, and focuses on the proposed solutions and their results. Demoireing is a difficult task of removing moire patterns from an image to reveal an underlying clean image. A new dataset, called LCDMoire was created for this challenge, and consists of 10,200 synthetically generated image pairs (moire and clean ground truth). The challenge was divided into 2 tracks. Track 1 targeted fidelity, measuring the ability of demoire methods to obtain a moire-free image compared with the ground truth, while Track 2 examined the perceptual quality of demoire methods. The tracks had 60 and 39 registered participants, respectively. A total of eight teams competed in the final testing phase. The entries span the current the state-of-the-art in the image demoireing problem.

eess.IV↗

arXiv2019

AIM 2019 Challenge on Image Demoireing: Dataset and Study

This paper introduces a novel dataset, called LCDMoire, which was created for the first-ever image demoireing challenge that was part of the Advances in Image Manipulation (AIM) workshop, held in conjunction with ICCV 2019. The dataset comprises 10,200 synthetically generated image pairs (consisting of an image degraded by moire and a clean ground truth image). In addition to describing the dataset and its creation, this paper also reviews the challenge tracks, competition, and results, the latter summarizing the current state-of-the-art on this dataset.

cs.CV↗

arXiv2019

AIM 2019 Challenge on Constrained Super-Resolution: Methods and Results

This paper reviews the AIM 2019 challenge on constrained example-based single image super-resolution with focus on proposed solutions and results. The challenge had 3 tracks. Taking the three main aspects (i.e., number of parameters, inference/running time, fidelity (PSNR)) of MSRResNet as the baseline, Track 1 aims to reduce the amount of parameters while being constrained to maintain or improve the running time and the PSNR result, Tracks 2 and 3 aim to optimize running time and PSNR result with constrain of the other two aspects, respectively. Each track had an average of 64 registered participants, and 12 teams submitted the final results. They gauge the state-of-the-art in single image super-resolution.

eess.IV↗

arXiv2019

Divide-and-Conquer Adversarial Learning for High-Resolution Image and Video Enhancement

This paper introduces a divide-and-conquer inspired adversarial learning (DACAL) approach for photo enhancement. The key idea is to decompose the photo enhancement process into hierarchically multiple sub-problems, which can be better conquered from bottom to up. On the top level, we propose a perception-based division to learn additive and multiplicative components, required to translate a low-quality image or video into its high-quality counterpart. On the intermediate level, we use a frequency-based division with generative adversarial network (GAN) to weakly supervise the photo enhancement process. On the lower level, we design a dimension-based division that enables the GAN model to better approximates the distribution distance on multiple independent one-dimensional data to train the GAN model. While considering all three hierarchies, we develop multiscale and recurrent training approaches to optimize the image and video enhancement process in a weakly-supervised manner. Both quantitative and qualitative results clearly demonstrate that the proposed DACAL achieves the state-of-the-art performance for high-resolution image and video enhancement.

eess.IV↗

arXiv2019

AI Benchmark: All About Deep Learning on Smartphones in 2019

The performance of mobile AI accelerators has been evolving rapidly in the past two years, nearly doubling with each new generation of SoCs. The current 4th generation of mobile NPUs is already approaching the results of CUDA-compatible Nvidia graphics cards presented not long ago, which together with the increased capabilities of mobile deep learning frameworks makes it possible to run complex and deep AI models on mobile devices. In this paper, we evaluate the performance and compare the results of all chipsets from Qualcomm, HiSilicon, Samsung, MediaTek and Unisoc that are providing hardware acceleration for AI inference. We also discuss the recent changes in the Android ML pipeline and provide an overview of the deployment of deep learning models on mobile devices. All numerical results provided in this paper can be found and are regularly updated on the official project website: http://ai-benchmark.com.

cs.PF↗

arXiv2019

Unsupervised Learning for Real-World Super-Resolution

Most current super-resolution methods rely on low and high resolution image pairs to train a network in a fully supervised manner. However, such image pairs are not available in real-world applications. Instead of directly addressing this problem, most works employ the popular bicubic downsampling strategy to artificially generate a corresponding low resolution image. Unfortunately, this strategy introduces significant artifacts, removing natural sensor noise and other real-world characteristics. Super-resolution networks trained on such bicubic images therefore struggle to generalize to natural images. In this work, we propose an unsupervised approach for image super-resolution. Given only unpaired data, we learn to invert the effects of bicubic downsampling in order to restore the natural image characteristics present in the data. This allows us to generate realistic image pairs, faithfully reflecting the distribution of real-world images. Our super-resolution network can therefore be trained with direct pixel-wise supervision in the high resolution domain, while robustly generalizing to real input. We demonstrate the effectiveness of our approach in quantitative and qualitative experiments.

eess.IV↗

arXiv2019

Extremely Weak Supervised Image-to-Image Translation for Semantic Segmentation

Recent advances in generative models and adversarial training have led to a flourishing image-to-image (I2I) translation literature. The current I2I translation approaches require training images from the two domains that are either all paired (supervised) or all unpaired (unsupervised). In practice, obtaining paired training data in sufficient quantities is often very costly and cumbersome. Therefore solutions that employ unpaired data, while less accurate, are largely preferred. In this paper, we aim to bridge the gap between supervised and unsupervised I2I translation, with application to semantic image segmentation. We build upon pix2pix and CycleGAN, state-of-the-art seminal I2I translation techniques. We propose a method to select (very few) paired training samples and achieve significant improvements in both supervised and unsupervised I2I translation settings over random selection. Further, we boost the performance by incorporating both (selected) paired and unpaired samples in the training process. Our experiments show that an extremely weak supervised I2I translation solution using only one paired training sample can achieve a quantitative performance much better than the unsupervised CycleGAN model, and comparable to that of the supervised pix2pix model trained on thousands of pairs.

cs.CV↗

arXiv2019

Efficient Video Super-Resolution through Recurrent Latent Space Propagation

With the recent trend for ultra high definition displays, the demand for high quality and efficient video super-resolution (VSR) has become more important than ever. Previous methods adopt complex motion compensation strategies to exploit temporal information when estimating the missing high frequency details. However, as the motion estimation problem is a highly challenging problem, inaccurate motion compensation may affect the performance of VSR algorithms. Furthermore, the complex motion compensation module may also introduce a heavy computational burden, which limits the application of these methods in real systems. In this paper, we propose an efficient recurrent latent space propagation (RLSP) algorithm for fast VSR. RLSP introduces high-dimensional latent states to propagate temporal information between frames in an implicit manner. Our experimental results show that RLSP is a highly efficient and effective method to deal with the VSR problem. We outperform current state-of-the-art method DUF with over 70x speed-up.

eess.IV↗

arXiv2019

SMIT: Stochastic Multi-Label Image-to-Image Translation

Cross-domain mapping has been a very active topic in recent years. Given one image, its main purpose is to translate it to the desired target domain, or multiple domains in the case of multiple labels. This problem is highly challenging due to three main reasons: (i) unpaired datasets, (ii) multiple attributes, and (iii) the multimodality (e.g., style) associated with the translation. Most of the existing state-of-the-art has focused only on two reasons, i.e. either on (i) and (ii), or (i) and (iii). In this work, we propose a joint framework (i, ii, iii) of diversity and multi-mapping image-to-image translations, using a single generator to conditionally produce countless and unique fake images that hold the underlying characteristics of the source image. Our system does not use style regularization, instead, it uses an embedding representation that we call domain embedding for both domain and style. Extensive experiments over different datasets demonstrate the effectiveness of our proposed approach in comparison with the state-of-the-art in both multi-label and multimodal problems. Additionally, our method is able to generalize under different scenarios: continuous style interpolation, continuous label interpolation, and fine-grained mapping. Code and pretrained models are available at https://github.com/BCV-Uniandes/SMIT.

cs.CV↗

arXiv2019

Generative Adversarial Networks for Extreme Learned Image Compression

We present a learned image compression system based on GANs, operating at extremely low bitrates. Our proposed framework combines an encoder, decoder/generator and a multi-scale discriminator, which we train jointly for a generative learned compression objective. The model synthesizes details it cannot afford to store, obtaining visually pleasing results at bitrates where previous methods fail and show strong artifacts. Furthermore, if a semantic label map of the original image is available, our method can fully synthesize unimportant regions in the decoded image such as streets and trees from the label map, proportionally reducing the storage cost. A user study confirms that for low bitrates, our approach is preferred to state-of-the-art methods, even when they use more than double the bits.

cs.CV↗

arXiv2019

Exemplar Guided Face Image Super-Resolution without Facial Landmarks

Nowadays, due to the ubiquitous visual media there are vast amounts of already available high-resolution (HR) face images. Therefore, for super-resolving a given very low-resolution (LR) face image of a person it is very likely to find another HR face image of the same person which can be used to guide the process. In this paper, we propose a convolutional neural network (CNN)-based solution, namely GWAInet, which applies super-resolution (SR) by a factor 8x on face images guided by another unconstrained HR face image of the same person with possible differences in age, expression, pose or size. GWAInet is trained in an adversarial generative manner to produce the desired high quality perceptual image results. The utilization of the HR guiding image is realized via the use of a warper subnetwork that aligns its contents to the input image and the use of a feature fusion chain for the extracted features from the warped guiding image and the input image. In training, the identity loss further helps in preserving the identity related features by minimizing the distance between the embedding vectors of SR and HR ground truth images. Contrary to the current state-of-the-art in face super-resolution, our method does not require facial landmark points for its training, which helps its robustness and allows it to produce fine details also for the surrounding face region in a uniform manner. Our method GWAInet produces photo-realistic images in upscaling factor 8x and outperforms state-of-the-art in quantitative terms and perceptual quality.

cs.CV↗

arXiv2019

Conditional Probability Models for Deep Image Compression

Deep Neural Networks trained as image auto-encoders have recently emerged as a promising direction for advancing the state-of-the-art in image compression. The key challenge in learning such networks is twofold: To deal with quantization, and to control the trade-off between reconstruction error (distortion) and entropy (rate) of the latent image representation. In this paper, we focus on the latter challenge and propose a new technique to navigate the rate-distortion trade-off for an image compression auto-encoder. The main idea is to directly model the entropy of the latent representation by using a context model: A 3D-CNN which learns a conditional probability model of the latent distribution of the auto-encoder. During training, the auto-encoder makes use of the context model to estimate the entropy of its representation, and the context model is concurrently updated to learn the dependencies between the symbols in the latent representation. Our experiments show that this approach, when measured in MS-SSIM, yields a state-of-the-art image compression system based on a simple convolutional auto-encoder.

cs.CV↗

arXiv2019

3D Appearance Super-Resolution with Deep Learning

We tackle the problem of retrieving high-resolution (HR) texture maps of objects that are captured from multiple view points. In the multi-view case, model-based super-resolution (SR) methods have been recently proved to recover high quality texture maps. On the other hand, the advent of deep learning-based methods has already a significant impact on the problem of video and image SR. Yet, a deep learning-based approach to super-resolve the appearance of 3D objects is still missing. The main limitation of exploiting the power of deep learning techniques in the multi-view case is the lack of data. We introduce a 3D appearance SR (3DASR) dataset based on the existing ETH3D [42], SyB3R [31], MiddleBury, and our Collection of 3D scenes from TUM [21], Fountain [51] and Relief [53]. We provide the high- and low-resolution texture maps, the 3D geometric model, images and projection matrices. We exploit the power of 2D learning-based SR methods and design networks suitable for the 3D multi-view case. We incorporate the geometric information by introducing normal maps and further improve the learning process. Experimental results demonstrate that our proposed networks successfully incorporate the 3D geometric information and super-resolve the texture maps.

cs.CV↗