In this paper, we propose a \textbf{Tr}ansformer-based RGB-D \textbf{e}gocentric \textbf{a}ction \textbf{r}ecognition framework, called Trear. It consists of two modules, inter-frame attention encoder and mutual-attentional fusion block. Instead of using optical flow or recurrent units, we adopt self-attention mechanism to model the temporal structure of the data from different modalities. Input frames are cropped randomly to mitigate the effect of the data redundancy. Features from each modality are interacted through the proposed fusion block and combined through a simple yet effective fusion operation to produce a joint RGB-D representation. Empirical experiments on two large egocentric RGB-D datasets, THU-READ and FPHA, and one small dataset, WCVS, have shown that the proposed method outperforms the state-of-the-art results by a large margin.
Temporal information is essential to learning effective policies with Reinforcement Learning (RL). However, current state-of-the-art RL algorithms either assume that such information is given as part of the state space or, when learning from pixels, use the simple heuristic of frame-stacking to implicitly capture temporal information present in the image observations. This heuristic is in contrast to the current paradigm in video classification architectures, which utilize explicit encodings of temporal information through methods such as optical flow and two-stream architectures to achieve state-of-the-art performance. Inspired by leading video classification architectures, we introduce the Flow of Latents for Reinforcement Learning (Flare), a network architecture for RL that explicitly encodes temporal information through latent vector differences. We show that Flare (i) recovers optimal performance in state-based RL without explicit access to the state velocity, solely with positional state information, (ii) achieves state-of-the-art performance on pixel-based challenging continuous control tasks within the DeepMind control benchmark suite, namely quadruped walk, hopper hop, finger turn hard, pendulum swing, and walker run, and is the most sample efficient model-free pixel-based RL algorithm, outperforming the prior model-free state-of-the-art by 1.9X and 1.5X on the 500k and 1M step benchmarks, respectively, and (iv), when augmented over rainbow DQN, outperforms this state-of-the-art level baseline on 5 of 8 challenging Atari games at 100M time step benchmark.
Establishing dense correspondences between a pair of images is an important and general problem. However, dense flow estimation is often inaccurate in the case of large displacements or homogeneous regions. For most applications and down-stream tasks, such as pose estimation, image manipulation, or 3D reconstruction, it is crucial to know when and where to trust the estimated matches. In this work, we aim to estimate a dense flow field relating two images, coupled with a robust pixel-wise confidence map indicating the reliability and accuracy of the prediction. We develop a flexible probabilistic approach that jointly learns the flow prediction and its uncertainty. In particular, we parametrize the predictive distribution as a constrained mixture model, ensuring better modelling of both accurate flow predictions and outliers. Moreover, we develop an architecture and training strategy tailored for robust and generalizable uncertainty prediction in the context of self-supervised training. Our approach obtains state-of-the-art results on multiple challenging geometric matching and optical flow datasets. We further validate the usefulness of our probabilistic confidence estimation for the task of pose estimation. Code and models are available at https://github.com/PruneTruong/PDCNet.
Video facial expression recognition is useful for many applications and received much interest lately. Although some solutions give really good results in a controlled environment (no occlusion), recognition in the presence of partial facial occlusion remains a challenging task. To handle occlusions, solutions based on the reconstruction of the occluded part of the face have been proposed. These solutions are mainly based on the texture or the geometry of the face. However, the similarity of the face movement between different persons doing the same expression seems to be a real asset for the reconstruction. In this paper we exploit this asset and propose a new solution based on an auto-encoder with skip connections to reconstruct the occluded part of the face in the optical flow domain. To the best of our knowledge, this is the first proposition to directly reconstruct the movement for facial expression recognition. We validated our approach in the controlled dataset CK+ on which different occlusions were generated. Our experiments show that the proposed method reduce significantly the gap, in terms of recognition accuracy, between occluded and non-occluded situations. We also compare our approach with existing state-of-the-art solutions. In order to lay the basis of a reproducible and fair comparison in the future, we also propose a new experimental protocol that includes occlusion generation and reconstruction evaluation.
With the development of cloud computing, the storage and processing of massive visual media data has gradually transferred to the cloud server. For example, if the intelligent video monitoring system cannot process a large amount of data locally, the data will be uploaded to the cloud. Therefore, how to process data in the cloud without exposing the original data has become an important research topic. We propose a single-server version of somewhat homomorphic encryption cryptosystem based on confused modulo projection theorem named CMP-SWHE, which allows the server to complete blind data processing without \emph{seeing} the effective information of user data. On the client side, the original data is encrypted by amplification, randomization, and setting confusing redundancy. Operating on the encrypted data on the server side is equivalent to operating on the original data. As an extension, we designed and implemented a blind computing scheme of accelerated version based on batch processing technology to improve efficiency. To make this algorithm easy to use, we also designed and implemented an efficient general blind computing library based on CMP-SWHE. We have applied this library to foreground extraction, optical flow tracking and object detection with satisfactory results, which are helpful for building smart cities. We also discuss how to extend the algorithm to deep learning applications. Compared with other homomorphic encryption cryptosystems and libraries, the results show that our method has obvious advantages in computing efficiency. Although our algorithm has some tiny errors ($10^{-6}$) when the data is too large, it is very efficient and practical, especially suitable for blind image and video processing.
A majority of methods for video frame interpolation compute bidirectional optical flow between adjacent frames of a video, followed by a suitable warping algorithm to generate the output frames. However, approaches relying on optical flow often fail to model occlusions and complex non-linear motions directly from the video and introduce additional bottlenecks unsuitable for widespread deployment. We address these limitations with FLAVR, a flexible and efficient architecture that uses 3D space-time convolutions to enable end-to-end learning and inference for video frame interpolation. Our method efficiently learns to reason about non-linear motions, complex occlusions and temporal abstractions, resulting in improved performance on video interpolation, while requiring no additional inputs in the form of optical flow or depth maps. Due to its simplicity, FLAVR can deliver 3x faster inference speed compared to the current most accurate method on multi-frame interpolation without losing interpolation accuracy. In addition, we evaluate FLAVR on a wide range of challenging settings and consistently demonstrate superior qualitative and quantitative results compared with prior methods on various popular benchmarks including Vimeo-90K, UCF101, DAVIS, Adobe, and GoPro. Finally, we demonstrate that FLAVR for video frame interpolation can serve as a useful self-supervised pretext task for action recognition, optical flow estimation, and motion magnification.
Speech-driven facial video generation has been a complex problem due to its multi-modal aspects namely audio and video domain. The audio comprises lots of underlying features such as expression, pitch, loudness, prosody(speaking style) and facial video has lots of variability in terms of head movement, eye blinks, lip synchronization and movements of various facial action units along with temporal smoothness. Synthesizing highly expressive facial videos from the audio input and static image is still a challenging task for generative adversarial networks. In this paper, we propose a multi-modal adaptive normalization(MAN) based architecture to synthesize a talking person video of arbitrary length using as input: an audio signal and a single image of a person. The architecture uses the multi-modal adaptive normalization, keypoint heatmap predictor, optical flow predictor and class activation map[58] based layers to learn movements of expressive facial components and hence generates a highly expressive talking-head video of the given person. The multi-modal adaptive normalization uses the various features of audio and video such as Mel spectrogram, pitch, energy from audio signals and predicted keypoint heatmap/optical flow and a single image to learn the respective affine parameters to generate highly expressive video. Experimental evaluation demonstrates superior performance of the proposed method as compared to Realistic Speech-Driven Facial Animation with GANs(RSDGAN) [53], Speech2Vid [10], and other approaches, on multiple quantitative metrics including: SSIM (structural similarity index), PSNR (peak signal to noise ratio), CPBD (image sharpness), WER(word error rate), blinks/sec and LMD(landmark distance). Further, qualitative evaluation and Online Turing tests demonstrate the efficacy of our approach.
In this paper, we consider the compressed video background subtraction problem that separates the background and foreground of a video from its compressed measurements. The background of a video usually lies in a low dimensional space and the foreground is usually sparse. More importantly, each video frame is a natural image that has textural patterns. By exploiting these properties, we develop a message passing algorithm termed offline denoising-based turbo message passing (DTMP). We show that these structural properties can be efficiently handled by the existing denoising techniques under the turbo message passing framework. We further extend the DTMP algorithm to the online scenario where the video data is collected in an online manner. The extension is based on the similarity/continuity between adjacent video frames. We adopt the optical flow method to refine the estimation of the foreground. We also adopt the sliding window based background estimation to reduce complexity. By exploiting the Gaussianity of messages, we develop the state evolution to characterize the per-iteration performance of offline and online DTMP. Comparing to the existing algorithms, DTMP can work at much lower compression rates, and can subtract the background successfully with a lower mean squared error and better visual quality for both offline and online compressed video background subtraction.
Independent Sign Language Recognition is a complex visual recognition problem that combines several challenging tasks of Computer Vision due to the necessity to exploit and fuse information from hand gestures, body features and facial expressions. While many state-of-the-art works have managed to deeply elaborate on these features independently, to the best of our knowledge, no work has adequately combined all three information channels to efficiently recognize Sign Language. In this work, we employ SMPL-X, a contemporary parametric model that enables joint extraction of 3D body shape, face and hands information from a single image. We use this holistic 3D reconstruction for SLR, demonstrating that it leads to higher accuracy than recognition from raw RGB images and their optical flow fed into the state-of-the-art I3D-type network for 3D action recognition and from 2D Openpose skeletons fed into a Recurrent Neural Network. Finally, a set of experiments on the body, face and hand features showed that neglecting any of these, significantly reduces the classification accuracy, proving the importance of jointly modeling body shape, facial expression and hand pose for Sign Language Recognition.
The paper addresses the problem of recognition of actions in video with low inter-class variability such as Table Tennis strokes. Two stream, "twin" convolutional neural networks are used with 3D convolutions both on RGB data and optical flow. Actions are recognized by classification of temporal windows. We introduce 3D attention modules and examine their impact on classification efficiency. In the context of the study of sportsmen performances, a corpus of the particular actions of table tennis strokes is considered. The use of attention blocks in the network speeds up the training step and improves the classification scores up to 5% with our twin model. We visualize the impact on the obtained features and notice correlation between attention and player movements and position. Score comparison of state-of-the-art action classification method and proposed approach with attentional blocks is performed on the corpus. Proposed model with attention blocks outperforms previous model without them and our baseline.
Short- or mid-term rainfall forecasting is a major task with several environmental applications such as agricultural management or flood risk monitoring. Existing data-driven approaches, especially deep learning models, have shown significant skill at this task, using only rainfall radar images as inputs. In order to determine whether using other meteorological parameters such as wind would improve forecasts, we trained a deep learning model on a fusion of rainfall radar images and wind velocity produced by a weather forecast model. The network was compared to a similar architecture trained only on radar data, to a basic persistence model and to an approach based on optical flow. Our network outperforms by 8% the F1-score calculated for the optical flow on moderate and higher rain events for forecasts at a horizon time of 30 min. Furthermore, it outperforms by 7% the same architecture trained using only rainfall radar images. Merging rain and wind data has also proven to stabilize the training process and enabled significant improvement especially on the difficult-to-predict high precipitation rainfalls.
As moving objects always draw more attention of human eyes, the temporal motive information is always exploited complementarily with spatial information to detect salient objects in videos. Although efficient tools such as optical flow have been proposed to extract temporal motive information, it often encounters difficulties when used for saliency detection due to the movement of camera or the partial movement of salient objects. In this paper, we investigate the complimentary roles of spatial and temporal information and propose a novel dynamic spatiotemporal network (DS-Net) for more effective fusion of spatiotemporal information. We construct a symmetric two-bypass network to explicitly extract spatial and temporal features. A dynamic weight generator (DWG) is designed to automatically learn the reliability of corresponding saliency branch. And a top-down cross attentive aggregation (CAA) procedure is designed so as to facilitate dynamic complementary aggregation of spatiotemporal features. Finally, the features are modified by spatial attention with the guidance of coarse saliency map and then go through decoder part for final saliency map. Experimental results on five benchmarks VOS, DAVIS, FBMS, SegTrack-v2, and ViSal demonstrate that the proposed method achieves superior performance than state-of-the-art algorithms. The source code is available at https://github.com/TJUMMG/DS-Net.
We study the problem of animating images by transferring spatio-temporal visual effects (such as melting) from a collection of videos. We tackle two primary challenges in visual effect transfer: 1) how to capture the effect we wish to distill; and 2) how to ensure that only the effect, rather than content or artistic style, is transferred from the source videos to the input image. To address the first challenge, we evaluate five loss functions; the most promising one encourages the generated animations to have similar optical flow and texture motions as the source videos. To address the second challenge, we only allow our model to move existing image pixels from the previous frame, rather than predicting unconstrained pixel values. This forces any visual effects to occur using the input image's pixels, preventing unwanted artistic style or content from the source video from appearing in the output. We evaluate our method in objective and subjective settings, and show interesting qualitative results which demonstrate objects undergoing atypical transformations, such as making a face melt or a deer bloom.
Patch-based approaches such as 3D block matching (BM3D) and non-local Bayes (NLB) are widely accepted filters for removing Gaussian noise from single-frame images. In this work, we propose three extensions for these filters when there exist multiple frames of the same scene. The first of them employs reference patches on every frame instead of a commonly used single reference frame method, thus utilizing the complete available information. The remaining two techniques use a separable spatio-temporal filter to reduce interactions between dissimilar regions, hence mitigating artifacts. In order to deal with non-registered datasets we combine all our extensions with robust optical flow computation. Two of our proposed multi-frame filters outperform existing extensions on most occasions by a significant margin while also being competitive with a state-of-the-art neural network-based technique. Moreover, one of these two strategies is the fastest among all due to its separable design.
We develop X-ray Multi-modal Intrinsic-Speckle-Tracking (MIST), a form of X-ray speckle-tracking that is able to recover both the position-dependent phase shift and the position-dependent small-angle X-ray scattering (SAXS) signal of a phase object. MIST is based on combining a Fokker-Planck description of paraxial X-ray optics, with an optical-flow formalism for X-ray speckle-tracking. Only two images need to be taken in the presence of the sample, corresponding to two different transverse positions of the speckle-generating membrane, in order to recover both the refractive and local-SAXS properties of the sample. Like the optical-flow X-ray phase-retrieval method which it generalises, the MIST method implicitly rather than explicitly tracks both the transverse motion and the diffusion of speckles that is induced by the presence of a sample. Application to X-ray synchrotron data shows the method to be efficient, rapid and stable.
The modeling of phenomenological structure is a crucial aspect in inverse imaging problems. One emerging modeling tool in computational imaging is the optimal transport framework. Its ability to model geometric displacements across an image's support gives it attractive qualities similar to those of optical flow methods which are effective at capturing visual motion, but are restricted to operate in significantly smaller state-spaces. Despite this advantage, two major drawbacks make it unsuitable for general deployment: (i) it suffers from exorbitant computational costs due to a quadratic optimization-variable complexity, and (ii) it has a mass-balancing assumption that limits applications with natural images. We tackle these issues simultaneously by proposing a novel formulation for an unbalanced optimal transport regularizer that has linear optimization-variable complexity. In addition, we present a general parallelizable proximal method for this regularizer, and demonstrate superior empirical performance on novel dynamical tracking applications in synthetic and real video.
Computed Tomography is a powerful imaging technique that allows non-destructive visualization of the interior of physical objects in different scientific areas. In traditional reconstruction techniques the object of interest is mostly considered to be static, which gives artefacts if the object is moving during the data acquisition. In this paper we present a method that, given only scan results of multiple successive scans, can estimate the motion and correct the CT-images for this motion assuming that the motion field is smooth over the complete domain using optical flow. The proposed method is validated on simulated scan data. The main contribution is that we show we can use the optical flow technique from imaging to correct CT-scan images for motion.
The pursuit of higher compression efficiency continuously drives the advances of video coding technologies. Fundamentally, we wish to find better "predictions" or "priors" that are reconstructed previously to remove the signal dependency efficiently and to accurately model the signal distribution for entropy coding. In this work, we propose a neural video compression framework, leveraging the spatial and temporal priors, independently and jointly to exploit the correlations in intra texture, optical flow based temporal motion and residuals. Spatial priors are generated using downscaled low-resolution features, while temporal priors (from previous reference frames and residuals) are captured using a convolutional neural network based long-short term memory (ConvLSTM) structure in a temporal recurrent fashion. All of these parts are connected and trained jointly towards the optimal rate-distortion performance. Compared with the High-Efficiency Video Coding (HEVC) Main Profile (MP), our method has demonstrated averaged 38% Bjontegaard-Delta Rate (BD-Rate) improvement using standard common test sequences, where the distortion is multi-scale structural similarity (MS-SSIM).
Detection of relative changes in circulating blood volume is important to guide resuscitation and manage variety of medical conditions including sepsis, trauma, dialysis and congestive heart failure. Recent studies have shown that estimates of circulating blood volume can be obtained from ultrasound imagery of the of the internal jugular vein (IJV). However, segmentation and tracking of the IJV is significantly influenced by speckle noise and shadowing which introduce uncertainty in the boundaries of the vessel. In this paper, we investigate the use of optical flow algorithms for segmentation and tracking of the IJV and show that the classical Lucas-Kanade (LK) algorithm provides the best performance among well-known flow tracking algorithms.
We propose an unsupervised vision-based system to estimate the joint configurations of the robot arm from a sequence of RGB or RGB-D images without knowing the model a priori, and then adapt it to the task of category-independent articulated object pose estimation. We combine a classical geometric formulation with deep learning and extend the use of epipolar constraint to multi-rigid-body systems to solve this task. Given a video sequence, the optical flow is estimated to get the pixel-wise dense correspondences. After that, the 6D pose is computed by a modified PnP algorithm. The key idea is to leverage the geometric constraints and the constraint between multiple frames. Furthermore, we build a synthetic dataset with different kinds of robots and multi-joint articulated objects for the research of vision-based robot control and robotic vision. We demonstrate the effectiveness of our method on three benchmark datasets and show that our method achieves higher accuracy than the state-of-the-art supervised methods in estimating joint angles of robot arms and articulated objects.