Adversarial attack methods have demonstrated the fragility of deep neural networks. Their imperceptible perturbations are frequently able fool classifiers into potentially dangerous misclassifications. We propose a novel way to interpret adversarial perturbations in terms of the effective input signal that classifiers actually use. Based on this, we apply specially trained autoencoders, referred to as S2SNets, as defense mechanism. They follow a two-stage training scheme: first unsupervised, followed by a fine-tuning of the decoder, using gradients from an existing classifier. S2SNets induce a shift in the distribution of gradients propagated through them, stripping them from class-dependent signal. We analyze their robustness against several white-box and gray-box scenarios on the large ImageNet dataset. Our approach reaches comparable resilience in white-box attack scenarios as other state-of-the-art defenses in gray-box scenarios. We further analyze the relationships of AlexNet, VGG 16, ResNet 50 and Inception v3 in adversarial space, and found that VGG 16 is the easiest to fool, while perturbations from ResNet 50 are the most transferable.
This paper strives for pixel-level segmentation of actors and their actions in video content. Different from existing works, which all learn to segment from a fixed vocabulary of actor and action pairs, we infer the segmentation from a natural language input sentence. This allows to distinguish between fine-grained actors in the same super-category, identify actor and action instances, and segment pairs that are outside of the actor and action vocabulary. We propose a fully-convolutional model for pixel-level actor and action segmentation using an encoder-decoder architecture optimized for video. To show the potential of actor and action video segmentation from a sentence, we extend two popular actor and action datasets with more than 7,500 natural language descriptions. Experiments demonstrate the quality of the sentence-guided segmentations, the generalization ability of our model, and its advantage for traditional actor and action segmentation compared to the state-of-the-art.
We propose a new end-to-end single image dehazing method, called Densely Connected Pyramid Dehazing Network (DCPDN), which can jointly learn the transmission map, atmospheric light and dehazing all together. The end-to-end learning is achieved by directly embedding the atmospheric scattering model into the network, thereby ensuring that the proposed method strictly follows the physics-driven scattering model for dehazing. Inspired by the dense network that can maximize the information flow along features from different levels, we propose a new edge-preserving densely connected encoder-decoder structure with multi-level pyramid pooling module for estimating the transmission map. This network is optimized using a newly introduced edge-preserving loss function. To further incorporate the mutual structural information between the estimated transmission map and the dehazed result, we propose a joint-discriminator based on generative adversarial network framework to decide whether the corresponding dehazed image and the estimated transmission map are real or fake. An ablation study is conducted to demonstrate the effectiveness of each module evaluated at both estimated transmission map and dehazed result. Extensive experiments demonstrate that the proposed method achieves significant improvements over the state-of-the-art methods. Code will be made available at: https://github.com/hezhangsprinter
Recognizing text from natural images is a hot research topic in computer vision due to its various applications. Despite the enduring research of several decades on optical character recognition (OCR), recognizing texts from natural images is still a challenging task. This is because scene texts are often in irregular (e.g. curved, arbitrarily-oriented or seriously distorted) arrangements, which have not yet been well addressed in the literature. Existing methods on text recognition mainly work with regular (horizontal and frontal) texts and cannot be trivially generalized to handle irregular texts. In this paper, we develop the arbitrary orientation network (AON) to directly capture the deep features of irregular texts, which are combined into an attention-based decoder to generate character sequence. The whole network can be trained end-to-end by using only images and word-level annotations. Extensive experiments on various benchmarks, including the CUTE80, SVT-Perspective, IIIT5k, SVT and ICDAR datasets, show that the proposed AON-based method achieves the-state-of-the-art performance in irregular datasets, and is comparable to major existing methods in regular datasets.
Neural sequence-to-sequence networks with attention have achieved remarkable performance for machine translation. One of the reasons for their effectiveness is their ability to capture relevant source-side contextual information at each time-step prediction through an attention mechanism. However, the target-side context is solely based on the sequence model which, in practice, is prone to a recency bias and lacks the ability to capture effectively non-sequential dependencies among words. To address this limitation, we propose a target-side-attentive residual recurrent network for decoding, where attention over previous words contributes directly to the prediction of the next word. The residual learning facilitates the flow of information from the distant past and is able to emphasize any of the previously translated words, hence it gains access to a wider context. The proposed model outperforms a neural MT baseline as well as a memory and self-attention network on three language pairs. The analysis of the attention learned by the decoder confirms that it emphasizes a wider context, and that it captures syntactic-like structures.
This paper considers the problem of single image depth estimation. The employment of convolutional neural networks (CNNs) has recently brought about significant advancements in the research of this problem. However, most existing methods suffer from loss of spatial resolution in the estimated depth maps; a typical symptom is distorted and blurry reconstruction of object boundaries. In this paper, toward more accurate estimation with a focus on depth maps with higher spatial resolution, we propose two improvements to existing approaches. One is about the strategy of fusing features extracted at different scales, for which we propose an improved network architecture consisting of four modules: an encoder, decoder, multi-scale feature fusion module, and refinement module. The other is about loss functions for measuring inference errors used in training. We show that three loss terms, which measure errors in depth, gradients and surface normals, respectively, contribute to improvement of accuracy in an complementary fashion. Experimental results show that these two improvements enable to attain higher accuracy than the current state-of-the-arts, which is given by finer resolution reconstruction, for example, with small objects and object boundaries.
An intelligent observer looks at the world and sees not only what is, but what is moving and what can be moved. In other words, the observer sees how the present state of the world can transform in the future. We propose a model that predicts future images by learning to represent the present state and its transformation given only a sequence of images. To do so, we introduce an architecture with a latent state composed of two components designed to capture (i) the present image state and (ii) the transformation between present and future states, respectively. We couple this latent state with a recurrent neural network (RNN) core that predicts future frames by transforming past states into future states by applying the accumulated state transformation with a learned operator. We describe how this model can be integrated into an encoder-decoder convolutional neural network (CNN) architecture that uses weighted residual connections to integrate representations of the past with representations of the future. Qualitatively, our approach generates image sequences that are stable and capture realistic motion over multiple predicted frames, without requiring adversarial training. Quantitatively, our method achieves prediction results comparable to state-of-the-art results on standard image prediction benchmarks (Moving MNIST, KTH, and UCF101).
This paper describes a novel energy-based probabilistic distribution that represents complex-valued data and explains how to apply it to direct feature extraction from complex-valued spectra. The proposed model, the complex-valued restricted Boltzmann machine (CRBM), is designed to deal with complex-valued visible units as an extension of the well-known restricted Boltzmann machine (RBM). Like the RBM, the CRBM learns the relationships between visible and hidden units without having connections between units in the same layer, which dramatically improves training efficiency by using Gibbs sampling or contrastive divergence (CD). Another important characteristic is that the CRBM also has connections between real and imaginary parts of each of the complex-valued visible units that help represent the data distribution in the complex domain. In speech signal processing, classification and generation features are often based on amplitude spectra (e.g., MFCC, cepstra, and mel-cepstra) even if they are calculated from complex spectra, and they ignore phase information. In contrast, the proposed feature extractor using the CRBM directly encodes the complex spectra (or another complex-valued representation of the complex spectra) into binary-valued latent features (hidden units). Since the visible-hidden connections are undirected, we can also recover (decode) the complex spectra from the latent features directly. Our speech coding experiments demonstrated that the CRBM outperformed other speech coding methods, such as methods using the conventional RBM, the mel-log spectrum approximate (MLSA) decoder, etc.
This paper addresses the problem of transparent object matting. Existing image matting approaches for transparent objects often require tedious capturing procedures and long processing time, which limit their practical use. In this paper, we first formulate transparent object matting as a refractive flow estimation problem. We then propose a deep learning framework, called TOM-Net, for learning the refractive flow. Our framework comprises two parts, namely a multi-scale encoder-decoder network for producing a coarse prediction, and a residual network for refinement. At test time, TOM-Net takes a single image as input, and outputs a matte (consisting of an object mask, an attenuation mask and a refractive flow field) in a fast feed-forward pass. As no off-the-shelf dataset is available for transparent object matting, we create a large-scale synthetic dataset consisting of 158K images of transparent objects rendered in front of images sampled from the Microsoft COCO dataset. We also collect a real dataset consisting of 876 samples using 14 transparent objects and 60 background images. Promising experimental results have been achieved on both synthetic and real data, which clearly demonstrate the effectiveness of our approach.
Chinese characters have a huge set of character categories, more than 20,000 and the number is still increasing as more and more novel characters continue being created. However, the enormous characters can be decomposed into a compact set of about 500 fundamental and structural radicals. This paper introduces a novel radical analysis network (RAN) to recognize printed Chinese characters by identifying radicals and analyzing two-dimensional spatial structures among them. The proposed RAN first extracts visual features from input by employing convolutional neural networks as an encoder. Then a decoder based on recurrent neural networks is employed, aiming at generating captions of Chinese characters by detecting radicals and two-dimensional structures through a spatial attention mechanism. The manner of treating a Chinese character as a composition of radicals rather than a single character class largely reduces the size of vocabulary and enables RAN to possess the ability of recognizing unseen Chinese character classes, namely zero-shot learning.
In this paper, the problem of describing visual contents of a video sequence with natural language is addressed. Unlike previous video captioning work mainly exploiting the cues of video contents to make a language description, we propose a reconstruction network (RecNet) with a novel encoder-decoder-reconstructor architecture, which leverages both the forward (video to sentence) and backward (sentence to video) flows for video captioning. Specifically, the encoder-decoder makes use of the forward flow to produce the sentence description based on the encoded video semantic features. Two types of reconstructors are customized to employ the backward flow and reproduce the video features based on the hidden state sequence generated by the decoder. The generation loss yielded by the encoder-decoder and the reconstruction loss introduced by the reconstructor are jointly drawn into training the proposed RecNet in an end-to-end fashion. Experimental results on benchmark datasets demonstrate that the proposed reconstructor can boost the encoder-decoder models and leads to significant gains in video caption accuracy.
This paper proposes an encoder-decoder network to disentangle shape features during 3D face reconstruction from single 2D images, such that the tasks of reconstructing accurate 3D face shapes and learning discriminative shape features for face recognition can be accomplished simultaneously. Unlike existing 3D face reconstruction methods, our proposed method directly regresses dense 3D face shapes from single 2D images, and tackles identity and residual (i.e., non-identity) components in 3D face shapes explicitly and separately based on a composite 3D face shape model with latent representations. We devise a training process for the proposed network with a joint loss measuring both face identification error and 3D face shape reconstruction error. To construct training data we develop a method for fitting 3D morphable model (3DMM) to multiple 2D images of a subject. Comprehensive experiments have been done on MICC, BU3DFE, LFW and YTF databases. The results show that our method expands the capacity of 3DMM for capturing discriminative shape features and facial detail, and thus outperforms existing methods both in 3D face reconstruction accuracy and in face recognition accuracy.
Most recent approaches use the sequence-to-sequence model for paraphrase generation. The existing sequence-to-sequence model tends to memorize the words and the patterns in the training dataset instead of learning the meaning of the words. Therefore, the generated sentences are often grammatically correct but semantically improper. In this work, we introduce a novel model based on the encoder-decoder framework, called Word Embedding Attention Network (WEAN). Our proposed model generates the words by querying distributed word representations (i.e. neural word embeddings), hoping to capturing the meaning of the according words. Following previous work, we evaluate our model on two paraphrase-oriented tasks, namely text simplification and short text abstractive summarization. Experimental results show that our model outperforms the sequence-to-sequence baseline by the BLEU score of 6.3 and 5.5 on two English text simplification datasets, and the ROUGE-2 F1 score of 5.7 on a Chinese summarization dataset. Moreover, our model achieves state-of-the-art performances on these three benchmark datasets.
Existing single view, 3D face reconstruction methods can produce beautifully detailed 3D results, but typically only for near frontal, unobstructed viewpoints. We describe a system designed to provide detailed 3D reconstructions of faces viewed under extreme conditions, out of plane rotations, and occlusions. Motivated by the concept of bump mapping, we propose a layered approach which decouples estimation of a global shape from its mid-level details (e.g., wrinkles). We estimate a coarse 3D face shape which acts as a foundation and then separately layer this foundation with details represented by a bump map. We show how a deep convolutional encoder-decoder can be used to estimate such bump maps. We further show how this approach naturally extends to generate plausible details for occluded facial regions. We test our approach and its components extensively, quantitatively demonstrating the invariance of our estimated facial details. We further provide numerous qualitative examples showing that our method produces detailed 3D face shapes in viewing conditions where existing state of the art often break down.
Neural style transfer has drawn broad attention in recent years. However, most existing methods aim to explicitly model the transformation between different styles, and the learned model is thus not generalizable to new styles. We here attempt to separate the representations for styles and contents, and propose a generalized style transfer network consisting of style encoder, content encoder, mixer and decoder. The style encoder and content encoder are used to extract the style and content factors from the style reference images and content reference images, respectively. The mixer employs a bilinear model to integrate the above two factors and finally feeds it into a decoder to generate images with target style and content. To separate the style features and content features, we leverage the conditional dependence of styles and contents given an image. During training, the encoder network learns to extract styles and contents from two sets of reference images in limited size, one with shared style and the other with shared content. This learning framework allows simultaneous style transfer among multiple styles and can be deemed as a special `multi-task' learning scenario. The encoders are expected to capture the underlying features for different styles and contents which is generalizable to new styles and contents. For validation, we applied the proposed algorithm to the Chinese Typeface transfer problem. Extensive experiment results on character generation have demonstrated the effectiveness and robustness of our method.
In the recent times, autoencoders, besides being used for compression, have been proven quite useful even for regenerating similar images or help in image denoising. They have also been explored for anomaly detection in a few cases. However, due to location invariance property of convolutional neural network, autoencoders tend to learn from or search for learned features in the complete image. This creates issues when all the items in the image are not equally important and their location matters. For such cases, a semi supervised solution - regional priority based autoencoder (RPAE) has been proposed. In this model, similar to object detection models, a region proposal network identifies the relevant areas in the images as belonging to one of the predefined categories and then those bounding boxes are fed into appropriate decoder based on the category they belong to. Finally, the error scores from all the decoders are combined based on their importance to provide total reconstruction error.
In this paper, we propose an efficient algorithm to directly restore a clear image from a hazy input. The proposed algorithm hinges on an end-to-end trainable neural network that consists of an encoder and a decoder. The encoder is exploited to capture the context of the derived input images, while the decoder is employed to estimate the contribution of each input to the final dehazed result using the learned representations attributed to the encoder. The constructed network adopts a novel fusion-based strategy which derives three inputs from an original hazy image by applying White Balance (WB), Contrast Enhancing (CE), and Gamma Correction (GC). We compute pixel-wise confidence maps based on the appearance differences between these different inputs to blend the information of the derived inputs and preserve the regions with pleasant visibility. The final dehazed image is yielded by gating the important features of the derived inputs. To train the network, we introduce a multi-scale approach such that the halo artifacts can be avoided. Extensive experimental results on both synthetic and real-world images demonstrate that the proposed algorithm performs favorably against the state-of-the-art algorithms.
Dense video captioning aims to generate text descriptions for all events in an untrimmed video. This involves both detecting and describing events. Therefore, all previous methods on dense video captioning tackle this problem by building two models, i.e. an event proposal and a captioning model, for these two sub-problems. The models are either trained separately or in alternation. This prevents direct influence of the language description to the event proposal, which is important for generating accurate descriptions. To address this problem, we propose an end-to-end transformer model for dense video captioning. The encoder encodes the video into appropriate representations. The proposal decoder decodes from the encoding with different anchors to form video event proposals. The captioning decoder employs a masking network to restrict its attention to the proposal event over the encoding feature. This masking network converts the event proposal to a differentiable mask, which ensures the consistency between the proposal and captioning during training. In addition, our model employs a self-attention mechanism, which enables the use of efficient non-recurrent structure during encoding and leads to performance improvements. We demonstrate the effectiveness of this end-to-end model on ActivityNet Captions and YouCookII datasets, where we achieved 10.12 and 6.58 METEOR score, respectively.
Dense video captioning is a newly emerging task that aims at both localizing and describing all events in a video. We identify and tackle two challenges on this task, namely, (1) how to utilize both past and future contexts for accurate event proposal predictions, and (2) how to construct informative input to the decoder for generating natural event descriptions. First, previous works predominantly generate temporal event proposals in the forward direction, which neglects future video context. We propose a bidirectional proposal method that effectively exploits both past and future contexts to make proposal predictions. Second, different events ending at (nearly) the same time are indistinguishable in the previous works, resulting in the same captions. We solve this problem by representing each event with an attentive fusion of hidden states from the proposal module and video contents (e.g., C3D features). We further propose a novel context gating mechanism to balance the contributions from the current event and its surrounding contexts dynamically. We empirically show that our attentively fused event representation is superior to the proposal hidden states or video contents alone. By coupling proposal and captioning modules into one unified framework, our model outperforms the state-of-the-arts on the ActivityNet Captions dataset with a relative gain of over 100% (Meteor score increases from 4.82 to 9.65).
Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-to-end deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at http://www.merl.com/research/license#FoldingNet