No file available
To read the file of this research,
you can request a copy directly from the authors.
Self-supervised Visual Feature Learning with Deep Neural Networks: A Survey
Preprints and early-stage research may not have been peer reviewed yet.
Abstract
Large-scale labeled data are generally required to train deep neural networks in order to obtain better performance in visual feature learning from images or videos for computer vision applications. To avoid extensive cost of collecting and annotating large-scale datasets, as a subset of unsupervised learning methods, self-supervised learning methods are proposed to learn general image and video features from large-scale unlabeled data without using any human-annotated labels. This paper provides an extensive review of deep learning-based self-supervised general visual feature learning methods from images or videos. First, the motivation, general pipeline, and terminologies of this field are described. Then the common deep neural network architectures that used for self-supervised learning are summarized. Next, the main components and evaluation metrics of self-supervised learning methods are reviewed followed by the commonly used image and video datasets and the existing self-supervised visual feature learning methods. Finally, quantitative performance comparisons of the reviewed methods on benchmark datasets are summarized and discussed for both image and video feature learning. At last, this paper is concluded and lists a set of promising future directions for self-supervised visual feature learning.
... The efficacy of these models relies on the multi-level representations of raw data via non-linear transformations, thereby enabling the encoding of semantic representations that can be leveraged for pattern recognition 32 . Nevertheless, the performance of deep learning models is highly correlated to the amount and quality of data available during the training phase 49 . The process of labeling data can be a slow, time-consuming, and biased endeavor. ...
Complex structures can be understood as compositions of smaller, more basic elements. The characterization of these structures requires an analysis of their constituents and their spatial configuration. Examples can be found in systems as diverse as galaxies, alloys, living tissues, cells, and even nanoparticles. In the latter field, the most challenging examples are those of subdivided particles and particle-based materials, due to the close proximity of their constituents. The characterization of such nanostructured materials is typically conducted through the utilization of micrographs. Despite the importance of micrograph analysis, the extraction of quantitative data is often constrained. The presented effort demonstrates the morphological characterization of subdivided particles utilizing a pre-trained artificial intelligence model. The results are validated using three types of nanoparticles: nanospheres, dumbbells, and trimers. The automated segmentation of whole particles, as well as their individual subdivisions, is investigated using the Segment Anything Model, which is based on a pre-trained neural network. The subdivisions of the particles are organized into sets, which presents a novel approach in this field. These sets collate data derived from a large ensemble of specific particle domains indicating to which particle each subdomain belongs. The arrangement of subdivisions into sets to characterize complex nanoparticles expands the information gathered from microscopy analysis. The presented method, which employs a pre-trained deep learning model, outperforms traditional techniques by circumventing systemic errors and human bias. It can effectively automate the analysis of particles, thereby providing more accurate and efficient results.
... , time-series analysis130 , video processing[131][132][133] , speech processing134 , and target tracking 135 for example. The most-cited papers involve medical research, solving problems such as accurate detection of tissue in monocular endoscopy136 , retinal disease diagnosis 137 , MRI parameter mapping or reconstructio138,139 , 3D medical-image analysis140 , and homography estimation 141 .However, we only found four papers dealing with self-supervised learning in RS using a small sample60,142- 144 . ...
Researchers and engineers have increasingly used Deep Learning (DL) for a variety of Remote Sensing (RS) tasks. However, data from local observations or via ground truth is often quite limited for training DL models, especially when these models represent key socio-environmental problems, such as the monitoring of extreme, destructive climate events, biodiversity, and sudden changes in ecosystem states. Such cases, also known as small data problems, pose significant methodological challenges. This review summarises these challenges in the RS domain and the possibility of using emerging DL techniques to overcome them. We show that the small data problem is a common challenge across disciplines and scales that results in poor model generalisability and transferability, yet this has not been investigated in a structured way. We first introduce ten emerging DL techniques: transfer learning, self-supervised learning, semi-supervised learning, few-shot learning, zero-shot learning, active learning, weakly supervised learning, multitask learning, process-aware learning, and ensemble learning; we also include a validation technique known as spatial k-fold cross validation. These techniques have shown promising potential in other scientific disciplines, but have been rarely applied in the RS domain. We also provide guidance on which learning technique to use in various cases, which helps to create a more methodologically robust DL application (and a greater number of them) that can be used to tackle socially important problems with limited data.
... A labeled dataset will always be a biased description of the visual world, due to having a fixed number of classes and being built with some systematic image collection process. Selfsupervised learning of visual representations (Jing and Tian, 2019) could serve as remedy for this issue. Self-supervised architectures could be fed with a stream completely unrelated images, collected from an arbitrary number of sources in a randomized way. ...
Classification performance based on ImageNet is the de-facto standard metric for CNN development. In this work we challenge the notion that CNN architecture design solely based on ImageNet leads to generally effective convolutional neural network (CNN) architectures that perform well on a diverse set of datasets and application domains. To this end, we investigate and ultimately improve ImageNet as a basis for deriving such architectures. We conduct an extensive empirical study for which we train 500 CNN architectures, sampled from the broad AnyNetX design space, on ImageNet as well as 8 additional well-known image classification benchmark datasets from a diverse array of application domains. We observe that the performances of the architectures are highly dataset dependent. Some datasets even exhibit a negative error correlation with ImageNet across all architectures. We show how to significantly increase these correlations by utilizing ImageNet subsets restricted to fewer classes. These contributions can have a profound impact on the way we design future CNN architectures and help alleviate the tilt we see currently in our community with respect to over-reliance on one dataset.
The Metaverse, a virtual reality (VR) space where users can interact with each other and digital objects, is rapidly becoming a reality. As this new world evolves, Artificial Intelligence (AI) is playing an increasingly important role in shaping its development. Integrating AI with emerging technologies in the Metaverse creates new possibilities for immersive experiences that were previously impossible. This paper explores how AI is integrated with technologies such as the Internet of Things, blockchain, Natural Language Processing, virtual reality, Augmented Reality, Mixed Reality, and Extended Reality. One potential benefit of using AI in the Metaverse is the ability to create personalized experiences for individual users, based on their behavior and preferences. Another potential benefit of using AI in the Metaverse is the ability to automate repetitive tasks, freeing up time and resources for more complex and creative endeavors. However, there are also challenges associated with using AI in the Metaverse, such as ensuring user privacy and addressing issues of bias and discrimination. By examining the potential benefits and challenges of using AI in the Metaverse, including ethical considerations, we can better prepare for this exciting new era of VR. This paper presents a comprehensive survey of AI and its integration with other emerging technologies in the Metaverse, as the Metaverse continues to evolve and grow, it will be important for developers and researchers to stay up to date with the latest developments in AI and emerging technologies to fully leverage their potential.
We present a principled approach to uncover the structure of visual data by solving a deep learning task coined visual permutation learning. The goal of this task is to find the permutation that recovers the structure of data from shuffled versions of it. In the case of natural images, this task boils down to recovering the original image from patches shuffled by an unknown permutation matrix. Permutation matrices are discrete, thereby posing difficulties for gradient-based optimization methods. To this end, we resort to a continuous approximation using doubly-stochastic matrices and formulate a novel bi-level optimization problem on such matrices that learns to recover the permutation. Unfortunately, such a scheme leads to expensive gradient computations. We circumvent this issue by further proposing a computationally cheap scheme for generating doubly stochastic matrices based on Sinkhorn iterations. To implement our approach we propose DeepPermNet, an end-to-end CNN model for this task. The utility of DeepPermNet is demonstrated on three challenging computer vision problems, namely, relative attributes learning, supervised learning-to-rank, and self-supervised representation learning. Our results show state-of-the-art performance on relative attributes learning, supervised learning-to-rank, and competitive results in the classification and segmentation tasks of the PASCAL VOC dataset for self-supervised learning.
In this paper we present ActiveStereoNet, the first deep learning solution for active stereo systems. Due to the lack of ground truth, our method is fully self-supervised, yet it produces precise depth with a subpixel precision of 1/30th of a pixel; it does not suffer from the common over-smoothing issues; it preserves the edges; and it explicitly handles occlusions. We introduce a novel reconstruction loss that is more robust to noise and texture-less patches, and is invariant to illumination changes. The proposed loss is optimized using a window-based cost aggregation with an adaptive support weight scheme. This cost aggregation is edge-preserving and smooths the loss function, which is key to allow the network to reach compelling results. Finally we show how the task of predicting invalid regions, such as occlusions, can be trained end-to-end without ground-truth. This component is crucial to reduce blur and particularly improves predictions along depth discontinuities. Extensive quantitatively and qualitatively evaluations on real and synthetic data demonstrate state of the art results in many challenging scenes.
Over the last years, deep convolutional neural networks (ConvNets) have transformed the field of computer vision thanks to their unparalleled capacity to learn high level semantic image features. However, in order to successfully learn those features, they usually require massive amounts of manually labeled data, which is both expensive and impractical to scale. Therefore, unsupervised semantic feature learning, i.e., learning without requiring manual annotation effort, is of crucial importance in order to successfully harvest the vast amount of visual data that are available today. In our work we propose to learn image features by training ConvNets to recognize the 2d rotation that is applied to the image that it gets as input. We demonstrate both qualitatively and quantitatively that this apparently simple task actually provides a very powerful supervisory signal for semantic feature learning. We exhaustively evaluate our method in various unsupervised feature learning benchmarks and we exhibit in all of them state-of-the-art performance. Specifically, our results on those benchmarks demonstrate dramatic improvements w.r.t. prior state-of-the-art approaches in unsupervised representation learning and thus significantly close the gap with supervised feature learning. For instance, in PASCAL VOC 2007 detection task our unsupervised pre-trained AlexNet model achieves the state-of-the-art (among unsupervised methods) mAP of 54.4% that is only 2.4 points lower from the supervised case. We get similarly striking results when we transfer our unsupervised learned features on various other tasks, such as ImageNet classification, PASCAL classification, PASCAL segmentation, and CIFAR-10 classification. The code and models of our paper will be published on: https://github.com/gidariss/FeatureLearningRotNet .
We present the Moments in Time Dataset, a large-scale human-annotated collection of one million short videos corresponding to dynamic events unfolding within three seconds. Modeling the spatial-audio-temporal dynamics even for actions occurring in 3 second videos poses many challenges: meaningful events do not include only people, but also objects, animals, and natural phenomena; visual and auditory events can be symmetrical or not in time ("opening" means "closing" in reverse order), and transient or sustained. We describe the annotation process of our dataset (each video is tagged with one action or activity label among 339 different classes), analyze its scale and diversity in comparison to other large-scale video datasets for action recognition, and report results of several baseline models addressing separately and jointly three modalities: spatial, temporal and auditory. The Moments in Time dataset designed to have a large coverage and diversity of events in both visual and auditory modalities, can serve as a new challenge to develop models that scale to the level of complexity and abstract reasoning that a human processes on a daily basis.
Generative Adversarial Networks (GANs) excel at creating realistic images with complex models for which maximum likelihood is infeasible. However, the convergence of GAN training has still not been proved. We propose a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions. TTUR has an individual learning rate for both the discriminator and the generator. Using the theory of stochastic approximation, we prove that the TTUR converges under mild assumptions to a stationary local Nash equilibrium. The convergence carries over to the popular Adam optimization, for which we prove that it follows the dynamics of a heavy ball with friction and thus prefers flat minima in the objective landscape. For the evaluation of the performance of GANs at image generation, we introduce the `Fréchet Inception Distance'' (FID) which captures the similarity of generated images to real ones better than the Inception Score. In experiments, TTUR improves learning for DCGANs and Improved Wasserstein GANs (WGAN-GP) outperforming conventional GAN training on CelebA, CIFAR-10, SVHN, LSUN Bedrooms, and the One Billion Word Benchmark. https://papers.nips.cc/paper/7240-gans-trained-by-a-two-time-scale-update-rule-converge-to-a-local-nash-equilibrium
The purpose of this study is to determine whether current video datasets have sufficient data for training very deep convolutional neural networks (CNNs) with spatio-temporal three-dimensional (3D) kernels. Recently, the performance levels of 3D CNNs in the field of action recognition have improved significantly. However, to date, conventional research has only explored relatively shallow 3D architectures. We examine the architectures of various 3D CNNs from relatively shallow to very deep ones on current video datasets. Based on the results of those experiments, the following conclusions could be obtained: (i) ResNet-18 training resulted in significant overfitting for UCF-101, HMDB-51, and ActivityNet but not for Kinetics. (ii) The Kinetics dataset has sufficient data for training of deep 3D CNNs, and enables training of up to 152 ResNets layers, interestingly similar to 2D ResNets on ImageNet. ResNeXt-101 achieved 78.4% average accuracy on the Kinetics test set. (iii) Kinetics pretrained simple 3D architectures outperforms complex 2D architectures, and the pretrained ResNeXt-101 achieved 94.5% and 70.2% on UCF-101 and HMDB-51, respectively. The use of 2D CNNs trained on ImageNet has produced significant progress in various tasks in image. We believe that using deep 3D CNNs together with Kinetics will retrace the successful history of 2D CNNs and ImageNet, and stimulate advances in computer vision for videos. The codes and pretrained models used in this study are publicly available. https://github.com/kenshohara/3D-ResNets-PyTorch
In the era of end-to-end deep learning, many advances in computer vision are driven by large amounts of labeled data. In the optical flow setting, however, obtaining dense per-pixel ground truth for real scenes is difficult and thus such data is rare. Therefore, recent end-to-end convolutional networks for optical flow rely on synthetic datasets for supervision, but the domain mismatch between training and test scenarios continues to be a challenge. Inspired by classical energy-based optical flow methods, we design an unsupervised loss based on occlusion-aware bidirectional flow estimation and the robust census transform to circumvent the need for ground truth flow. On the KITTI benchmarks, our unsupervised approach outperforms previous unsupervised deep networks by a large margin, and is even more accurate than similar supervised methods trained on synthetic datasets alone. By optionally fine-tuning on the KITTI training data, our method achieves competitive optical flow accuracy on the KITTI 2012 and 2015 benchmarks, thus in addition enabling generic pre-training of supervised networks for datasets with limited amounts of ground truth.
Automatically describing the content of an image is a fundamental problem in
artificial intelligence that connects computer vision and natural language
processing. In this paper, we present a generative model based on a deep
recurrent architecture that combines recent advances in computer vision and
machine translation and that can be used to generate natural sentences
describing an image. The model is trained to maximize the likelihood of the
target description sentence given the training image. Experiments on several
datasets show the accuracy of the model and the fluency of the language it
learns solely from image descriptions. Our model is often quite accurate, which
we verify both qualitatively and quantitatively. For instance, while the
current state-of-the-art BLEU score (the higher the better) on the Pascal
dataset is 25, our approach yields 59, to be compared to human performance
around 69. We also show BLEU score improvements on Flickr30k, from 55 to 66,
and on SBU, from 19 to 27.
Recent work has shown that convolutional networks can be substantially deeper, more accurate and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper we embrace this observation and introduce the Dense Convolutional Network (DenseNet), where each layer is directly connected to every other layer in a feed-forward fashion. Whereas traditional convolutional networks with L layers have L connections, one between each layer and its subsequent layer (treating the input as layer 0), our network has L(L+1)/2 direct connections. For each layer, the feature maps of all preceding layers are treated as separate inputs whereas its own feature maps are passed on as inputs to all subsequent layers. Our proposed connectivity pattern has several compelling advantages: it alleviates the vanishing gradient problem and strengthens feature propagation; despite the increase in connections, it encourages feature reuse and leads to a substantial reduction of parameters; its models tend to generalize surprisingly well. We evaluate our proposed architecture on five highly competitive object recognition benchmark tasks. The DenseNet obtains significant improvements over the state-of-the-art on all five of them (e.g., yielding 3.74% test error on CIFAR-10, 19.25% on CIFAR-100 and 1.59% on SVHN).
Self-supervised tasks such as colorization, inpainting and zigsaw puzzle have been utilized for visual representation learning for still images, when the number of labeled images is limited or absent at all. Recently, this worthwhile stream of study extends to video domain where the cost of human labeling is even more expensive. However, the most of existing methods are still based on 2D CNN architectures that can not directly capture spatio-temporal information for video applications. In this paper, we introduce a new self-supervised task called as Space-Time Cubic Puzzles to train 3D CNNs using large scale video dataset. This task requires a network to arrange permuted 3D spatio-temporal crops. By completing Space-Time Cubic Puzzles, the network learns both spatial appearance and temporal relation of video frames, which is our final goal. In experiments, we demonstrate that our learned 3D representation is well transferred to action recognition tasks, and outperforms state-of-the-art 2D CNN-based competitors on UCF101 and HMDB51 datasets.
We propose a novel method for learning convolutional neural image representations without manual supervision. We use motion cues in the form of optical-flow, to supervise representations of static images. The obvious approach of training a network to predict flow from a single image can be needlessly difficult due to intrinsic ambiguities in this prediction task. We instead propose a much simpler learning goal: embed pixels such that the similarity between their embeddings matches that between their optical-flow vectors. At test time, the learned deep network can be used without access to video or flow information and transferred to tasks such as image classification, detection, and segmentation. Our method, which significantly simplifies previous attempts at using motion for self-supervision, achieves state-of-the-art results in self-supervision using motion cues, and is overall state of the art in self-supervised pre-training for semantic image segmentation, as demonstrated on standard benchmarks.
The thud of a bouncing ball, the onset of speech as lips open -- when visual and audio events occur together, it suggests that there might be a common, underlying event that produced both signals. In this paper, we argue that the visual and audio components of a video signal should be modeled jointly using a fused multisensory representation. We propose to learn such a representation in a self-supervised way, by training a neural network to predict whether video frames and audio are temporally aligned. We use this learned representation for three applications: (a) sound source localization, i.e. visualizing the source of sound in a video; (b) audio-visual action recognition; and (c) on/off-screen audio source separation, e.g. removing the off-screen translator's voice from a foreign official's speech. Code, models, and video results are available on our webpage: http://andrewowens.com/multisensory
We propose GeoNet, a jointly unsupervised learning framework for monocular depth, optical flow and ego-motion estimation from videos. The three components are coupled by the nature of 3D scene geometry, jointly learned by our framework in an end-to-end manner. Specifically, geometric relationships are extracted over the predictions of individual modules and then combined as an image reconstruction loss, reasoning about static and dynamic scene parts separately. Furthermore, we propose an adaptive geometric consistency loss to increase robustness towards outliers and non-Lambertian regions, which resolves occlusions and texture ambiguities effectively. Experimentation on the KITTI driving dataset reveals that our scheme achieves state-of-the-art results in all of the three tasks, performing better than previously unsupervised methods and comparably with supervised ones.
In this paper, we propose an efficient and straightforward approach, video you only look once (VideoYOLO), to capture the overall temporal dynamics from an entire video in a single process for action recognition. It remains an open question for action recognition on how to deal with the temporal dimension in videos. Existing methods subdivide a whole video into either individual frames or short clips and consequently have to process these fractions multiple times. A post process is then used to aggregate the partial dynamic cues to implicitly infer the whole temporal information. On the contrary, in VideoYOLO, we first generate a proxy video by selecting a subset of frames to roughly reserve the overall temporal dynamics presented in the original video. A 3D convolutional neural network (3D-CNN) is employed to learn the overall temporal characteristics from the proxy video and predict action category in a single process. Our proposed method is extremely fast. VideoYOLO-32 is able to process 36 videos per second that is 10 times and 7 times faster than prior 2D-CNN (Two-stream (Simonyan and Zisserman, 2014)) and 3D-CNN (C3D (Tran et al., 2015)) based models, respectively, while still achieves superior or comparable classification accuracies on the benchmark datasets, UCF101 and HMDB51.
As an agent moves through the world, the apparent motion of scene elements is (usually) inversely proportional to their depth. It is natural for a learning agent to associate image patterns with the magnitude of their displacement over time: as the agent moves, far away mountains don't move much; nearby trees move a lot. This natural relationship between the appearance of objects and their motion is a rich source of information about the world. In this work, we train a deep network, using fully automatic supervision, to predict relative scene depth from single images. The depth training images are automatically derived from simple videos of cars moving through a scene, using classic depth from motion techniques, and no human provided labels. We show that this pretext task of predicting depth from a single image induces features in the network that result in large improvements in a set of downstream tasks including semantic segmentation, joint road segmentation and car detection, and monocular (absolute) depth estimation, over a network trained from scratch. In particular, our pre-trained model outperforms an ImageNet counterpart for the monocular depth estimation task. Unlike work that analyzes video paired with additional information about direction of motion, our agent learns from "raw egomotion" video recorded from cars moving through the world. Unlike methods that require videos of moving objects, we neither depend on, nor are disrupted by, moving objects in the video. Indeed, we can benefit from predicting depth in the videos associated with various downstream tasks, showing that we can adapt to new scenes in an unsupervised manner to improve performance. By doing so, we achieve consistently better results over several different urban scene understanding tasks, obtaining results that are competitive with state-of-the-art method for monocular depth estimation.
In human learning, it is common to use multiple sources of information jointly. However, most existing feature learning approaches learn from only a single task. In this paper, we propose a novel multi-task deep network to learn generalizable high-level visual representations. Since multi-task learning requires annotations for multiple properties of the same training instance, we look to synthetic images to train our network. To overcome the domain difference between real and synthetic data, we employ an unsupervised feature space domain adaptation method based on adversarial learning. Given an input synthetic RGB image, our network simultaneously predicts its surface normal, depth, and instance contour, while also minimizing the feature space domain differences between real and synthetic data. Through extensive experiments, we demonstrate that our network learns more transferable representations compared to single-task baselines. Our learned representation produces state-of-the-art transfer learning results on PASCAL VOC 2007 classification and 2012 detection.
We study the question of feature sets for robust visual object recognition, adopting linear SVM based human detection as a test case. After reviewing existing edge and gradient based descriptors, we show experimentally that grids of Histograms of Oriented Gradient (HOG) descriptors significantly outperform existing feature sets for human detection. We study the influence of each stage of the computation on performance, concluding that fine-scale gradients, fine orientation binning, relatively coarse spatial binning, and high-quality local contrast normalization in overlapping descriptor blocks are all important for good results. The new approach gives near-perfect separation on the original MIT pedestrian database, so we introduce a more challenging dataset containing over 1800 annotated human images with a large range of pose variations and backgrounds.
A great deal of research has focused on algorithms for learning features from un- labeled data. Indeed, much progress has been made on benchmark datasets like NORB and CIFAR by employing increasingly complex unsupervised learning al- gorithms and deep models. In this paper, however, we show that several very sim- ple factors, such as the number of hidden nodes in the model, may be as important to achieving high performance as the choice of learning algorithm or the depth of the model. Specifically, we will apply several off-the-shelf feature learning al- gorithms (sparse auto-encoders, sparse RBMs and K-means clustering, Gaussian mixtures) to NORB and CIFAR datasets using only single-layer networks. We then present a detailed analysis of the effect of changes in the model setup: the receptive field size, number of hidden nodes (features), the step-size (stride) be- tween extracted features, and the effect of whitening. Our results show that large numbers of hidden nodes and dense feature extraction are as critical to achieving high performance as the choice of algorithm itselfso critical, in fact, that when these parameters are pushed to their limits, we are able to achieve state-of-the- art performance on both CIFAR and NORB using only a single layer of features. More surprisingly, our best performance is based on K-means clustering, which is extremely fast, has no hyper-parameters to tune beyond the model structure it- self, and is very easy implement. Despite the simplicity of our system, we achieve performance beyond all previously published results on the CIFAR-10 and NORB datasets (79.6% and 97.0% accuracy respectively).
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to 1/2 everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
Learning to predict future images from a video sequence involves the construction of an internal representation that models the image evolution accurately, and therefore, to some degree, its content and dynamics. This is why pixel-space video prediction is viewed as a promising avenue for unsupervised feature learning. In this work, we train a convolutional network to generate future frames given an input sequence. To deal with the inherently blurry predictions obtained from the standard Mean Squared Error (MSE) loss function, we propose three different and complementary feature learning strategies: a multi-scale architecture, an adversarial training method, and an image gradient difference loss function. We compare our predictions to different published results based on recurrent neural networks on the UCF101 dataset.
We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 dif- ferent classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0% which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully-connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implemen- tation of the convolution operation. To reduce overfitting in the fully-connected layers we employed a recently-developed regularization method called dropout that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry
Given a grayscale photograph as input, this paper attacks the problem of hallucinating a backslashem plausible color version of the photograph. This problem is clearly underconstrained, so previous approaches have either relied on significant user interaction or resulted in desaturated colorizations. We propose a fully automatic approach that produces vibrant and realistic colorizations. We embrace the underlying uncertainty of the problem by posing it as a classification task and explore using class-rebalancing at training time to increase the diversity of colors in the result. The system is implemented as a feed-forward operation in a CNN at test time and is trained on over a million color images. We evaluate our algorithm using a "colorization Turing test", asking human subjects to choose between a generated and ground truth color image. Our method successfully fools humans 20backslash% of the time, significantly higher than previous methods.
State-of-the-art object detection networks depend on region proposal algorithms to hypothesize object locations. Advances like SPPnet [7] and Fast R-CNN [5] have reduced the running time of these detection networks, exposing region pro-posal computation as a bottleneck. In this work, we introduce a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals. An RPN is a fully-convolutional network that simultaneously predicts object bounds and objectness scores at each position. RPNs are trained end-to-end to generate high-quality region proposals, which are used by Fast R-CNN for detection. With a simple alternating optimization, RPN and Fast R-CNN can be trained to share convolu-tional features. For the very deep VGG-16 model [18], our detection system has a frame rate of 5fps (including all steps) on a GPU, while achieving state-of-the-art object detection accuracy on PASCAL VOC 2007 (73.2% mAP) and 2012 (70.4% mAP) using 300 proposals per image. The code will be released.
We introduce a novel method for representation learning that uses an artificial supervision signal based on counting visual primitives. This supervision signal is obtained from an equivariance relation, which does not require any manual annotation. We relate transformations of images to transformations of the representations. More specifically, we look for the representation that satisfies such relation rather than the transformations that match a given representation. In this paper, we use two image transformations in the context of counting: scaling and tiling. The first transformation exploits the fact that the number of visual primitives should be invariant to scale. The second transformation allows us to equate the total number of visual primitives in each tile to that in the whole image. These two transformations are combined in one constraint and used to train a neural network with a contrastive loss. The proposed task produces representations that perform on par or exceed the state of the art in transfer learning benchmarks.
We present an unsupervised representation learning approach using videos without semantic labels. We leverage the temporal coherence as a supervisory signal by formulating representation learning as a sequence sorting task. We take temporally shuffled frames (i.e., in non-chronological order) as inputs and train a convolutional neural network to sort the shuffled sequences. Similar to comparison-based sorting algorithms, we propose to extract features from all frame pairs and aggregate them to predict the correct order. As sorting shuffled image sequence requires an understanding of the statistical temporal structure of images, training with such a proxy task allows us to learn rich and generalizable visual representation. We validate the effectiveness of the learned representation using our method as pre-training on high-level recognition problems. The experimental results show that our method compares favorably against state-of-the-art methods on action recognition, image classification and object detection tasks.
Future frame prediction in videos is a promising avenue for unsupervised video representation learning. Video frames are naturally generated by the inherent pixel flows from preceding frames based on the appearance and motion dynamics in the video. However, existing methods focus on directly hallucinating pixel values, resulting in blurry predictions. In this paper, we develop a dual motion Generative Adversarial Net (GAN) architecture, which learns to explicitly enforce future-frame predictions to be consistent with the pixel-wise flows in the video through a dual-learning mechanism. The primal future-frame prediction and dual future-flow prediction form a closed loop, generating informative feedback signals to each other for better video prediction. To make both synthesized future frames and flows indistinguishable from reality, a dual adversarial training method is proposed to ensure that the future-flow prediction is able to help infer realistic future-frames, while the future-frame prediction in turn leads to realistic optical flows. Our dual motion GAN also handles natural motion uncertainty in different pixel locations with a new probabilistic motion encoder, which is based on variational autoencoders. Extensive experiments demonstrate that the proposed dual motion GAN significantly outperforms state-of-the-art approaches on synthesizing new video frames and predicting future flows. Our model generalizes well across diverse visual scenes and shows superiority in unsupervised video representation learning.
We present a novel approach for image completion that results in images that are both locally and globally consistent. With a fully-convolutional neural network, we can complete images of arbitrary resolutions by filling-in missing regions of any shape. To train this image completion network to be consistent, we use global and local context discriminators that are trained to distinguish real images from completed ones. The global discriminator looks at the entire image to assess if it is coherent as a whole, while the local discriminator looks only at a small area centered at the completed region to ensure the local consistency of the generated patches. The image completion network is then trained to fool the both context discriminator networks, which requires it to generate images that are indistinguishable from real ones with regard to overall consistency as well as in details. We show that our approach can be used to complete a wide variety of scenes. Furthermore, in contrast with the patch-based approaches such as PatchMatch, our approach can generate fragments that do not appear elsewhere in the image, which allows us to naturally complete the images of objects with familiar and highly specific structures, such as faces.