How to learn discriminative video representation from unlabeled videos is challenging but crucial for video analysis. The latest attempts seek to learn a representation model by predicting the appearance contents in the masked regions. However, simply masking and recovering appearance contents may not be sufficient to model temporal clues as the appearance contents can be easily reconstructed from a single frame. To overcome this limitation, we present Masked Motion Encoding (MME), a new pre-training paradigm that reconstructs both appearance and motion information to explore temporal clues. In MME, we focus on addressing two critical challenges to improve the representation performance: 1) how to well represent the possible long-term motion across multiple frames; and 2) how to obtain fine-grained temporal clues from sparsely sampled videos. Motivated by the fact that human is able to recognize an action by tracking objects' position changes and shape changes, we propose to reconstruct a motion trajectory that represents these two kinds of change in the masked regions. Besides, given the sparse video input, we enforce the model to reconstruct dense motion trajectories in both spatial and temporal dimensions. Pre-trained with our MME paradigm, the model is able to anticipate long-term and fine-grained motion details. Code is available at https://github.com/XinyuSun/MME.
We present a robust theoretical framework for the characterisation and manipulation of optical flow, i.e 2D vector fields, in the context of their use in motion estimation algorithms and beyond. The definition of two frames of reference guides the mathematical derivation of flow field application, inversion, evaluation, and composition operations. This structured approach is then used as the foundation for an implementation in Python 3, with the fully differentiable PyTorch version oflibpytorch supporting back-propagation as required for deep learning. We verify the flow composition method empirically and provide a working example for its application to optical flow ground truth in synthetic training data creation. All code is publicly available.
Indirect Time-of-Flight (iToF) cameras are a widespread type of 3D sensor, which perform multiple captures to obtain depth values of the captured scene. While recent approaches to correct iToF depths achieve high performance when removing multi-path-interference and sensor noise, little research has been done to tackle motion artifacts. In this work we propose a training algorithm, which allows to supervise Optical Flow (OF) networks directly on the reconstructed depth, without the need of having ground truth flows. We demonstrate that this approach enables the training of OF networks to align raw iToF measurements and compensate motion artifacts in the iToF depth images. The approach is evaluated for both single- and multi-frequency sensors as well as multi-tap sensors, and is able to outperform other motion compensation techniques.
A single rolling-shutter (RS) image may be viewed as a row-wise combination of a sequence of global-shutter (GS) images captured by a (virtual) moving GS camera within the exposure duration. Although RS cameras are widely used, the RS effect causes obvious image distortion especially in the presence of fast camera motion, hindering downstream computer vision tasks. In this paper, we propose to invert the RS image capture mechanism, i.e., recovering a continuous high framerate GS video from two time-consecutive RS frames. We call this task the RS temporal super-resolution (RSSR) problem. The RSSR is a very challenging task, and to our knowledge, no practical solution exists to date. This paper presents a novel deep-learning based solution. By leveraging the multi-view geometry relationship of the RS imaging process, our learning-based framework successfully achieves high framerate GS generation. Specifically, three novel contributions can be identified: (i) novel formulations for bidirectional RS undistortion flows under constant velocity as well as constant acceleration motion model. (ii) a simple linear scaling operation, which bridges the RS undistortion flow and regular optical flow. (iii) a new mutual conversion scheme between varying RS undistortion flows that correspond to different scanlines. Our method also exploits the underlying spatial-temporal geometric relationships within a deep learning framework, where no additional supervision is required beyond the necessary middle-scanline GS image. Building upon these contributions, we represent the very first rolling-shutter temporal super-resolution deep-network that is able to recover high framerate GS videos from just two RS frames. Extensive experimental results on both synthetic and real data show that our proposed method can produce high-quality GS image sequences with rich details, outperforming the state-of-the-art methods.
Event cameras provide an advantage over traditional frame-based cameras when capturing fast-moving objects without a motion blur. They achieve this by recording changes in light intensity (known as events), thus allowing them to operate at a much higher frequency and making them suitable for capturing motions in a highly dynamic scene. Many recent studies have proposed methods to train neural networks (NNs) for predicting optical flow from events. However, they often rely on a spatio-temporal representation constructed from events over a fixed interval, such as 10Hz used in training on the DSEC dataset. This limitation restricts the flow prediction to the same interval (10Hz) whereas the fast speed of event cameras, which can operate up to 3kHz, has not been effectively utilized. In this work, we show that a temporally dense flow estimation at 100Hz can be achieved by treating the flow estimation as a sequential problem using two different variants of recurrent networks - Long-short term memory (LSTM) and spiking neural network (SNN). First, We utilize the NN model constructed similar to the popular EV-FlowNet but with LSTM layers to demonstrate the efficiency of our training method. The model not only produces 10x more frequent optical flow than the existing ones, but the estimated flows also have 13% lower errors than predictions from the baseline EV-FlowNet. Second, we construct an EV-FlowNet SNN but with leaky integrate and fire neurons to efficiently capture the temporal dynamics. We found that simple inherent recurrent dynamics of SNN lead to significant parameter reduction compared to the LSTM model. In addition, because of its event-driven computation, the spiking model is estimated to consume only 1.5% energy of the LSTM model, highlighting the efficiency of SNN in processing events and the potential for achieving temporally dense flow.
When applied to autonomous vehicle (AV) settings, action recognition can enhance an environment model's situational awareness. This is especially prevalent in scenarios where traditional geometric descriptions and heuristics in AVs are insufficient. However, action recognition has traditionally been studied for humans, and its limited adaptability to noisy, un-clipped, un-pampered, raw RGB data has limited its application in other fields. To push for the advancement and adoption of action recognition into AVs, this work proposes a novel two-stage action recognition system, termed RALACs. RALACs formulates the problem of action recognition for road scenes, and bridges the gap between it and the established field of human action recognition. This work shows how attention layers can be useful for encoding the relations across agents, and stresses how such a scheme can be class-agnostic. Furthermore, to address the dynamic nature of agents on the road, RALACs constructs a novel approach to adapting Region of Interest (ROI) Alignment to agent tracks for downstream action classification. Finally, our scheme also considers the problem of active agent detection, and utilizes a novel application of fusing optical flow maps to discern relevant agents in a road scene. We show that our proposed scheme can outperform the baseline on the ICCV2021 Road Challenge dataset and by deploying it on a real vehicle platform, we provide preliminary insight to the usefulness of action recognition in decision making.
Video frame interpolation (VFI) is a fundamental vision task that aims to synthesize several frames between two consecutive original video images. Most algorithms aim to accomplish VFI by using only keyframes, which is an ill-posed problem since the keyframes usually do not yield any accurate precision about the trajectories of the objects in the scene. On the other hand, event-based cameras provide more precise information between the keyframes of a video. Some recent state-of-the-art event-based methods approach this problem by utilizing event data for better optical flow estimation to interpolate for video frame by warping. Nonetheless, those methods heavily suffer from the ghosting effect. On the other hand, some of kernel-based VFI methods that only use frames as input, have shown that deformable convolutions, when backed up with transformers, can be a reliable way of dealing with long-range dependencies. We propose event-based video frame interpolation with attention (E-VFIA), as a lightweight kernel-based method. E-VFIA fuses event information with standard video frames by deformable convolutions to generate high quality interpolated frames. The proposed method represents events with high temporal resolution and uses a multi-head self-attention mechanism to better encode event-based information, while being less vulnerable to blurring and ghosting artifacts; thus, generating crispier frames. The simulation results show that the proposed technique outperforms current state-of-the-art methods (both frame and event-based) with a significantly smaller model size.
We propose SF2SE3, a novel approach to estimate scene dynamics in form of a segmentation into independently moving rigid objects and their SE(3)-motions. SF2SE3 operates on two consecutive stereo or RGB-D images. First, noisy scene flow is obtained by application of existing optical flow and depth estimation algorithms. SF2SE3 then iteratively (1) samples pixel sets to compute SE(3)-motion proposals, and (2) selects the best SE(3)-motion proposal with respect to a maximum coverage formulation. Finally, objects are formed by assigning pixels uniquely to the selected SE(3)-motions based on consistency with the input scene flow and spatial proximity. The main novelties are a more informed strategy for the sampling of motion proposals and a maximum coverage formulation for the proposal selection. We conduct evaluations on multiple datasets regarding application of SF2SE3 for scene flow estimation, object segmentation and visual odometry. SF2SE3 performs on par with the state of the art for scene flow estimation and is more accurate for segmentation and odometry.
Detecting dangerous traffic agents in videos captured by vehicle-mounted dashboard cameras (dashcams) is essential to facilitate safe navigation in a complex environment. Accident-related videos are just a minor portion of the driving video big data, and the transient pre-accident processes are highly dynamic and complex. Besides, risky and non-risky traffic agents can be similar in their appearance. These make risky object localization in the driving video particularly challenging. To this end, this paper proposes an attention-guided multistream feature fusion network (AM-Net) to localize dangerous traffic agents from dashcam videos. Two Gated Recurrent Unit (GRU) networks use object bounding box and optical flow features extracted from consecutive video frames to capture spatio-temporal cues for distinguishing dangerous traffic agents. An attention module coupled with the GRUs learns to attend to the traffic agents relevant to an accident. Fusing the two streams of features, AM-Net predicts the riskiness scores of traffic agents in the video. In supporting this study, the paper also introduces a benchmark dataset called Risky Object Localization (ROL). The dataset contains spatial, temporal, and categorical annotations with the accident, object, and scene-level attributes. The proposed AM-Net achieves a promising performance of 85.73% AUC on the ROL dataset. Meanwhile, the AM-Net outperforms current state-of-the-art for video anomaly detection by 6.3% AUC on the DoTA dataset. A thorough ablation study further reveals AM-Net's merits by evaluating the contributions of its different components.
Unsupervised video object segmentation aims to segment a target object in the video without a ground truth mask in the initial frame. This challenging task requires extracting features for the most salient common objects within a video sequence. This difficulty can be solved by using motion information such as optical flow, but using only the information between adjacent frames results in poor connectivity between distant frames and poor performance. To solve this problem, we propose a novel prototype memory network architecture. The proposed model effectively extracts the RGB and motion information by extracting superpixel-based component prototypes from the input RGB images and optical flow maps. In addition, the model scores the usefulness of the component prototypes in each frame based on a self-learning algorithm and adaptively stores the most useful prototypes in memory and discards obsolete prototypes. We use the prototypes in the memory bank to predict the next query frames mask, which enhances the association between distant frames to help with accurate mask prediction. Our method is evaluated on three datasets, achieving state-of-the-art performance. We prove the effectiveness of the proposed model with various ablation studies.
Unsupervised video object segmentation (VOS) aims to detect the most salient object in a video sequence at the pixel level. In unsupervised VOS, most state-of-the-art methods leverage motion cues obtained from optical flow maps in addition to appearance cues to exploit the property that salient objects usually have distinctive movements compared to the background. However, as they are overly dependent on motion cues, which may be unreliable in some cases, they cannot achieve stable prediction. To reduce this motion dependency of existing two-stream VOS methods, we propose a novel motion-as-option network that optionally utilizes motion cues. Additionally, to fully exploit the property of the proposed network that motion is not always required, we introduce a collaborative network learning strategy. On all the public benchmark datasets, our proposed network affords state-of-the-art performance with real-time inference speed.
Masked visual modeling (MVM) has been recently proven effective for visual pre-training. While similar reconstructive objectives on video inputs (e.g., masked frame modeling) have been explored in video-language (VidL) pre-training, previous studies fail to find a truly effective MVM strategy that can largely benefit the downstream performance. In this work, we systematically examine the potential of MVM in the context of VidL learning. Specifically, we base our study on a fully end-to-end VIdeO-LanguagE Transformer (VIOLET), where the supervision from MVM training can be backpropagated to the video pixel space. In total, eight different reconstructive targets of MVM are explored, from low-level pixel values and oriented gradients to high-level depth maps, optical flow, discrete visual tokens, and latent visual features. We conduct comprehensive experiments and provide insights into the factors leading to effective MVM training, resulting in an enhanced model VIOLETv2. Empirically, we show VIOLETv2 pre-trained with MVM objective achieves notable improvements on 13 VidL benchmarks, ranging from video question answering, video captioning, to text-to-video retrieval.
Weakly supervised video object localization (WSVOL) allows locating object in videos using only global video tags such as object class. State-of-art methods rely on multiple independent stages, where initial spatio-temporal proposals are generated using visual and motion cues, then prominent objects are identified and refined. Localization is done by solving an optimization problem over one or more videos, and video tags are typically used for video clustering. This requires a model per-video or per-class making for costly inference. Moreover, localized regions are not necessary discriminant because of unsupervised motion methods like optical flow, or because video tags are discarded from optimization. In this paper, we leverage the successful class activation mapping (CAM) methods, designed for WSOL based on still images. A new Temporal CAM (TCAM) method is introduced to train a discriminant deep learning (DL) model to exploit spatio-temporal information in videos, using an aggregation mechanism, called CAM-Temporal Max Pooling (CAM-TMP), over consecutive CAMs. In particular, activations of regions of interest (ROIs) are collected from CAMs produced by a pretrained CNN classifier to build pixel-wise pseudo-labels for training the DL model. In addition, a global unsupervised size constraint, and local constraint such as CRF are used to yield more accurate CAMs. Inference over single independent frames allows parallel processing of a clip of frames, and real-time localization. Extensive experiments on two challenging YouTube-Objects datasets for unconstrained videos, indicate that CAM methods (trained on independent frames) can yield decent localization accuracy. Our proposed TCAM method achieves a new state-of-art in WSVOL accuracy, and visual results suggest that it can be adapted for subsequent tasks like visual object tracking and detection. Code is publicly available.
Latent space exploration is a technique that discovers interpretable latent directions and manipulates latent codes to edit various attributes in images generated by generative adversarial networks (GANs). However, in previous work, spatial control is limited to simple transformations (e.g., translation and rotation), and it is laborious to identify appropriate latent directions and adjust their parameters. In this paper, we tackle the problem of editing the StyleGAN image layout by annotating the image directly. To do so, we propose an interactive framework for manipulating latent codes in accordance with the user inputs. In our framework, the user annotates a StyleGAN image with locations they want to move or not and specifies a movement direction by mouse dragging. From these user inputs and initial latent codes, our latent transformer based on a transformer encoder-decoder architecture estimates the output latent codes, which are fed to the StyleGAN generator to obtain a result image. To train our latent transformer, we utilize synthetic data and pseudo-user inputs generated by off-the-shelf StyleGAN and optical flow models, without manual supervision. Quantitative and qualitative evaluations demonstrate the effectiveness of our method over existing methods.
In this work, we propose a motion robust and high-speed detection pipeline which better leverages the event data. First, we design an event stream representation called temporal active focus (TAF), which efficiently utilizes the spatial-temporal asynchronous event stream, constructing event tensors robust to object motions. Then, we propose a module called the bifurcated folding module (BFM), which encodes the rich temporal information in the TAF tensor at the input layer of the detector. Following this, we design a high-speed lightweight detector called agile event detector (AED) plus a simple but effective data augmentation method, to enhance the detection accuracy and reduce the model's parameter. Experiments on two typical real-scene event camera object detection datasets show that our method is competitive in terms of accuracy, efficiency, and the number of parameters. By classifying objects into multiple motion levels based on the optical flow density metric, we further illustrated the robustness of our method for objects with different velocities relative to the camera. The codes and trained models are available at https://github.com/HarmoniaLeo/FRLW-EvD .
Learned video compression methods have gained a variety of interest in the video coding community since they have matched or even exceeded the rate-distortion (RD) performance of traditional video codecs. However, many current learning-based methods are dedicated to utilizing short-range temporal information, thus limiting their performance. In this paper, we focus on exploiting the unique characteristics of video content and further exploring temporal information to enhance compression performance. Specifically, for long-range temporal information exploitation, we propose temporal prior that can update continuously within the group of pictures (GOP) during inference. In that case temporal prior contains valuable temporal information of all decoded images within the current GOP. As for short-range temporal information, we propose a progressive guided motion compensation to achieve robust and effective compensation. In detail, we design a hierarchical structure to achieve multi-scale compensation. More importantly, we use optical flow guidance to generate pixel offsets between feature maps at each scale, and the compensation results at each scale will be used to guide the following scale's compensation. Sufficient experimental results demonstrate that our method can obtain better RD performance than state-of-the-art video compression approaches. The code is publicly available on: https://github.com/Huairui/LSTVC.
Dense depth and pose estimation is a vital prerequisite for various video applications. Traditional solutions suffer from the robustness of sparse feature tracking and insufficient camera baselines in videos. Therefore, recent methods utilize learning-based optical flow and depth prior to estimate dense depth. However, previous works require heavy computation time or yield sub-optimal depth results. We present GCVD, a globally consistent method for learning-based video structure from motion (SfM) in this paper. GCVD integrates a compact pose graph into the CNN-based optimization to achieve globally consistent estimation from an effective keyframe selection mechanism. It can improve the robustness of learning-based methods with flow-guided keyframes and well-established depth prior. Experimental results show that GCVD outperforms the state-of-the-art methods on both depth and pose estimation. Besides, the runtime experiments reveal that it provides strong efficiency in both short- and long-term videos with global consistency provided.
Domain shifts, such as appearance changes, are a key challenge in real-world applications of activity recognition models, which range from assistive robotics and smart homes to driver observation in intelligent vehicles. For example, while simulations are an excellent way of economical data collection, a Synthetic-to-Real domain shift leads to a > 60% drop in accuracy when recognizing activities of Daily Living (ADLs). We tackle this challenge and introduce an activity domain generation framework which creates novel ADL appearances (novel domains) from different existing activity modalities (source domains) inferred from video training data. Our framework computes human poses, heatmaps of body joints, and optical flow maps and uses them alongside the original RGB videos to learn the essence of source domains in order to generate completely new ADL domains. The model is optimized by maximizing the distance between the existing source appearances and the generated novel appearances while ensuring that the semantics of an activity is preserved through an additional classification loss. While source data multimodality is an important concept in this design, our setup does not rely on multi-sensor setups, (i.e., all source modalities are inferred from a single video only.) The newly created activity domains are then integrated in the training of the ADL classification networks, resulting in models far less susceptible to changes in data distributions. Extensive experiments on the Synthetic-to-Real benchmark Sims4Action demonstrate the potential of the domain generation paradigm for cross-domain ADL recognition, setting new state-of-the-art results. Our code is publicly available at https://github.com/Zrrr1997/syn2real_DG
Previous methods based on 3DCNN, convLSTM, or optical flow have achieved great success in video salient object detection (VSOD). However, they still suffer from high computational costs or poor quality of the generated saliency maps. To solve these problems, we design a space-time memory (STM)-based network, which extracts useful temporal information of the current frame from adjacent frames as the temporal branch of VSOD. Furthermore, previous methods only considered single-frame prediction without temporal association. As a result, the model may not focus on the temporal information sufficiently. Thus, we initially introduce object motion prediction between inter-frame into VSOD. Our model follows standard encoder--decoder architecture. In the encoding stage, we generate high-level temporal features by using high-level features from the current and its adjacent frames. This approach is more efficient than the optical flow-based methods. In the decoding stage, we propose an effective fusion strategy for spatial and temporal branches. The semantic information of the high-level features is used to fuse the object details in the low-level features, and then the spatiotemporal features are obtained step by step to reconstruct the saliency maps. Moreover, inspired by the boundary supervision commonly used in image salient object detection (ISOD), we design a motion-aware loss for predicting object boundary motion and simultaneously perform multitask learning for VSOD and object motion prediction, which can further facilitate the model to extract spatiotemporal features accurately and maintain the object integrity. Extensive experiments on several datasets demonstrated the effectiveness of our method and can achieve state-of-the-art metrics on some datasets. The proposed model does not require optical flow or other preprocessing, and can reach a speed of nearly 100 FPS during inference.
Temporal consistency is the key challenge of video depth estimation. Previous works are based on additional optical flow or camera poses, which is time-consuming. By contrast, we derive consistency with less information. Since videos inherently exist with heavy temporal redundancy, a missing frame could be recovered from neighboring ones. Inspired by this, we propose the frame masking network (FMNet), a spatial-temporal transformer network predicting the depth of masked frames based on their neighboring frames. By reconstructing masked temporal features, the FMNet can learn intrinsic inter-frame correlations, which leads to consistency. Compared with prior arts, experimental results demonstrate that our approach achieves comparable spatial accuracy and higher temporal consistency without any additional information. Our work provides a new perspective on consistent video depth estimation. Our official project page is https://github.com/RaymondWang987/FMNet.