Abstract

Abstract— Learning computational models for visual attention (saliency estimation) is an effort to inch machines/robots closer to human visual cognitive abilities. Data-driven efforts have dominated the landscape since the introduction of deep neural network architectures. In deep learning research, the choices in architecture design are often empirical and frequently lead to more complex models than necessary. The complexity, in turn, hinders the application requirements. In this paper, we identify four key components of saliency models, i.e., input features, multi-level integration, readout architecture, and loss functions. We review the existing state of the art models on these four components and propose novel and simpler alternatives. As a result, we propose two novel end-to-end architectures called SimpleNet and MDNSal, which are neater, minimal, more interpretable and achieve state of the art performance on public saliency benchmarks. SimpleNet is an optimized encoder-decoder architecture and brings notable performance gains on the SALICON dataset (the largest saliency benchmark). MDNSal is a parametric model that directly predicts parameters of a GMM distribution and is aimed to bring more interpretability to the prediction maps. The proposed saliency models can be inferred at 25fps, making them suitable for realtime applications. Code and pre-trained models are available at https://github.com/samyak0210/saliency.

Abstract

Abstract— Detecting small obstacles on the road is critical for autonomous driving. In this paper, we present a method to reliably detect such obstacles through a multi-modal framework of sparse LiDAR(VLP-16) and Monocular vision. LiDAR is employed to provide additional context in the form of confidence maps to monocular segmentation networks. We show significant performance gains when the context is fed as an additional input to monocular semantic segmentation frameworks. We further present a new semantic segmentation dataset to the community, comprising of over 3000 image frames with corresponding LiDAR observations. The images come with pixel-wise annotations of three classes off-road, road, and small obstacle. We stress that precise calibration between LiDAR and camera is crucial for this task and thus propose a novel Hausdorff distance based calibration refinement method over extrinsic parameters. As a first benchmark over this dataset, we report our results with 73 % instance detection up to a distance of 50 meters on challenging scenarios. Qualitatively by showcasing accurate segmentation of obstacles less than 15 cms at 50m depth and quantitatively through favourable comparisons vis a vis prior art, we vindicate the method’s efficacy. Our project and dataset is hosted at https://small-obstacle-dataset.github.io/

Abstract

Colorization is a complex task of selecting a combination of colors and arriving at an appropriate spatial arrangement of the colors in an image. In this paper, we propose a novel approach for automatic colorization of graphic arts like graphic patterns, info-graphics and cartoons. Our approach uses the artist’s colored graphics as a reference to color a template image. We also propose a retrieval system for selecting a relevant reference image corresponding to the given template from a dataset of reference images colored by different artists. Finally, we formulate the problem of colorization as a optimal graph matching problem over color groups in the reference and the template image. We demonstrate results on a variety of coloring tasks and evaluate our model through multiple perceptual studies. The studies show that the results generated through our model are significantly preferred by the participants over other automatic colorization methods.

Abstract

We present GAZED– eye GAZ-guided EDiting for videos captured by a solitary, static, wide-angle and high-resolution camera. Eye-gaze has been effectively employed in computational applications as a cue to capture interesting scene content; we employ gaze as a proxy to select shots for inclusion in the edited video. Given the original video, scene content and user eye-gaze tracks are combined to generate an edited video comprising of cinematically valid actor shots and shot transitions to generate an aesthetic and vivid representation of the original narrative. We model cinematic video editing as an energy minimization problem over shot selection, whose constraints capture cinematographic editing conventions. Gazed scene locations primarily determine the shots constituting the edited video. Effectiveness of GAZED against multiple competing methods is demonstrated via a psychophysical study involving 12 users and twelve performance videos. Professional video recordings of stage performances are typically created by employing skilled camera operators, who record the performance from multiple viewpoints. These multi-camera feeds, termed rushes, are then edited together to portray an eloquent story intended to maximize viewer engagement. Generating professional edits of stage performances is both difficult and challenging. Firstly, maneuvering cameras during a live performance is difficult even for experts as there is no option of retake upon error, and camera viewpoints are limited as the use of large supporting equipment (trolley, crane .) is infeasible. Secondly, manual video editing is an extremely slow and tedious process and leverages the experience of skilled editors. Overall, the need for (i) a professional camera crew, (ii) multiple cameras and supporting equipment, and (iii) expert editors escalates the process complexity and costs. Consequently, most production houses employ a large field-of-view static camera, placed far enough to capture the entire stage. This approach is widespread as it is simple to implement, and also captures the entire scene. Such static visualizations are apt for archival purposes; however, they are often unsuccessful at captivating attention when presented to the target audience. While conveying the overall context, the distant camera feed fails to bring out vivid scene details like close-up faces, character emotions and actions, and ensuing interactions which are critical for cinematic storytelling. GAZED denotes an end-to-end pipeline to generate an aesthetically edited video from a single static, wide-angle stage recording. This is inspired by prior work [GRG14], which describes how a plural camera crew can be replaced by a single highresolution static camera, and multiple virtual camera shots or rushes generated by simulating several virtual pan/tilt/zoom cameras to focus on actors and actions within the original recording. In this work, we demonstrate that the multiple rushes can be automatically edited by leveraging user eye gaze information, by modeling (virtual) shot selection as a discrete optimization problem. Eye-gaze represents an inherent guiding factor for video editing, as eyes are sensitive to interesting scene events [RKH∗ 09, SSSM14] that need to be vividly presented in the edited video. The objective critical for video editing and the key contribution of our work is to decide which shot (or rush) needs to be selected to describe each frame of the edited video. The shot selection problem is modeled as an optimization, which incorporates gaze information along with other cost terms that model cinematic editing principles. Gazed scene locations are utilized to define gaze potentials, which measure the importance of the different shots to choose from. Gaze potentials are then combined with other terms that model cinematic principles like avoiding jump cuts (which produce jarring shot transitions), rhythm (pace of shot transitioning), avoiding transient shots . The optimization is solved using dynamic programming. [MKSG20] refers to the detailed full article.

Abstract

Multi-object tracking has seen a lot of progress recently, albeit with substantial annotation costs for developing better and larger labeled datasets. In this work, we remove the need for annotated datasets by proposing an unsupervised re-identification network, thus sidestepping the labeling costs entirely, required for training. Given unlabeled videos, our proposed method (SimpleReID) first generates tracking labels using SORT [ 3] and trains a ReID network to predict the generated labels using crossentropy loss. We demonstrate that SimpleReID performs substantially better than simpler alternatives, and we recover the full performance of its supervised counterpart consistently across diverse tracking frameworks. The observations are unusual because unsupervised ReID is not expected to excel in crowded scenarios with occlusions, and drastic viewpoint changes. By incorporating our unsupervised SimpleReID with CenterTrack trained on augmented still images, we establish a new state-of-the-art performance on popular datasets like MOT16/17 without using tracking supervision, beating current best (CenterTrack) by 0.2-0.3 MOTA and 4.4-4.8 IDF1 scores. We further provide evidence for limited scope for improvement in IDF1 scores beyond our unsupervised ReID in the studied settings. Our investigation suggests reconsideration towards more sophisticated, supervised, end-to-end trackers [56 , 5] by showing promise in simpler unsupervised alternatives.

Abstract

Localizing natural language phrases in images is a challenging problem that requires joint understanding of both the textual and visual modalities. In the unsupervised setting, lack of supervisory signals exacerbate this difficulty. In this paper, we propose a novel framework for unsupervised visual grounding which uses concept learning as a proxy task to obtain self-supervision. The intuition behind this idea is to encourage the model to localize to regions which can explain some semantic property in the data, in our case, the property being the presence of a concept in a set of images. We present thorough quantitative and qualitative experiments to demonstrate the efficacy of our approach and show a 5.6% improvement over the current state of the art on Visual Genome dataset, a 5.8% improvement on the ReferItGame dataset and comparable to state-of-art performance on the Flickr30k dataset.

Abstract

Monocular head pose estimation requires learning a model that computes the intrinsic Euler angles for pose (yaw, pitch, roll) from an input image of human face. Annotating ground truth head pose angles for images in the wild is difficult and requires ad-hoc fitting procedures (which provides only coarse and approximate annotations). This highlights the need for approaches which can train on data captured in controlled environment and generalize on the images in the wild (with varying appearance and illumination of the face). Most present day deep learning approaches which learn a regression function directly on the input images fail to do so. To this end, we propose to use a higher level representation to regress the head pose while using deep learning architectures. More specifically, we use the uncertainty maps in the form of 2D soft localization heatmap images over five facial keypoints, namely left ear, right ear, left eye, right eye and nose, and pass them through an convolutional neural network to regress the head-pose. We show head pose estimation results on two challenging benchmarks BIWI and AFLW and our approach surpasses the state of the art on both the datasets. Index Terms— Image analysis, Pose estimation

Abstract

Abstract— We propose a novel pipeline that blends encodings from natural language and 3D semantic maps obtained from visual imagery to generate local trajectories that are executed by a low-level controller. The pipeline precludes the need for a prior registered map through a local waypoint generator neural network. The waypoint generator network (WGN) maps semantics and natural language encodings (NLE) to local waypoints. A local planner then generates a trajectory from the ego location of the vehicle (an outdoor car in this case) to these locally generated waypoints while a low-level controller executes these plans faithfully. The efficacy of the pipeline is verified in the CARLA simulator environment as well as on local semantic maps built from real-world KITTI dataset. In both these environments (simulated and real-world) we show the ability of the WGN to generate waypoints accurately by mapping NLE of varying sequence lengths and levels of complexity. We compare with baseline approaches and show significant performance gain over them. And finally, we show real implementations on our electric car verifying that the pipeline lends itself to practical and tangible realizations in uncontrolled outdoor settings. In loop execution of the proposed pipeline that involves repetitive invocations of the network is critical for any such language-based navigation framework. This effort successfully accomplishes this thereby bypassing the need for prior metric maps or strategies for metric level localization during traversal.

Abstract

Localizing natural language phrases in images is a challenging problem that requires joint understanding of both the textual and visual modalities. In the unsupervised setting, lack of supervisory signals exacerbate this difficulty. In this paper, we propose a novel framework for unsupervised visual grounding which uses concept learning as a proxy task to obtain self-supervision. The intuition behind this idea is to encourage the model to localize to regions which can explain some semantic property in the data, in our case, the property being the presence of a concept in a set of images. We present thorough quantitative and qualitative experiments to demonstrate the efficacy of our approach and show a 5.6% improvement over the current state of the art on Visual Genome dataset, a 5.8% improvement on the ReferItGame dataset and comparable to state-of-art performance on the Flickr30k dataset

Abstract

In this paper, we propose a novel method to register football broadcast video frames on the static top view model of the playing surface. The proposed method is fully automatic in contrast to the current state of the art which requires manual initialization of point correspondences between the image and the static model. Automatic registration using existing approaches has been difficult due to the lack of sufficient point correspondences. We investigate an alternate approach exploiting the edge information from the line markings on the field. We formulate the registration problem as a nearest neighbour search over a synthetically generated dictionary of edge map and homography pairs. The synthetic dictionary generation allows us to exhaustively cover a wide variety of camera angles and positions and reduce this problem to a minimal per-frame edge map matching procedure. We show that the per-frame results can be improved in videos using an optimization framework for temporal camera stabilization. We demonstrate the efficacy of our approach by presenting extensive results on a dataset collected from matches of football World Cup 2014.