Scene understanding under low-light conditions is a challenging problem. This is due to the small number of photons captured by the camera and the resulting low signal-to-noise ratio (SNR). Single-photon cameras (SPCs) are an emerging sensing modality that are capable of capturing images with high sensitivity. Despite having minimal read-noise, images captured by SPCs in photon-starved conditions still suffer from strong shot noise, preventing reliable scene inference. We propose photon scale-space a collection of high-SNR images spanning a wide range of photons-per-pixel (PPP) levels (but same scene content) as guides to train inference model on low photon flux images. We develop training techniques that push images with different illumination levels closer to each other in feature representation space. The key idea is that having a spectrum of different brightness levels during training enables effective guidance, and increases robustness to shot noise even in extreme noise cases. Based on the proposed approach, we demonstrate, via simulations and real experiments with a SPAD camera, high-performance on various inference tasks such as image classification and monocular depth estimation under ultra low-light, down to < 1 PPP.
Causality knowledge is vital to building robust AI systems. Deep learning models often perform poorly on tasks that require causal reasoning, which is often derived using some form of commonsense knowledge not immediately available in the input but implicitly inferred by humans. Prior work has unraveled spurious observational biases that models fall prey to in the absence of causality. While language representation models preserve contextual knowledge within learned embeddings, they do not factor in causal relationships during training. By blending causal relationships with the input features to an existing model that performs visual cognition tasks (such as scene understanding, video captioning, video question-answering, etc.), better performance can be achieved owing to the insight causal relationships bring about. Recently, several models have been proposed that have tackled the task of mining causal data from either the visual or textual modality. However, there does not exist widespread research that mines causal relationships by juxtaposing the visual and language modalities. While images offer a rich and easy-to-process resource for us to mine causality knowledge from, videos are denser and consist of naturally time-ordered events. Also, textual information offers details that could be implicit in videos. We propose iReason, a framework that infers visual-semantic commonsense knowledge using both videos and natural language captions. Furthermore, iReason's architecture integrates a causal rationalization module to aid the process of interpretability, error analysis and bias detection. We demonstrate the effectiveness of iReason using a two-pronged comparative analysis with language representation learning models (BERT, GPT-2) as well as current state-of-the-art multimodal causality models.
In recent years, many semantic segmentation methods have been proposed to predict label of pixels in the scene. In general, we measure area prediction errors or boundary prediction errors for comparing methods. However, there is no intuitive evaluation metric that evaluates both aspects. In this work, we propose a new evaluation measure called weighted Intersection over Union (wIoU) for semantic segmentation. First, it builds a weight map generated from a boundary distance map, allowing weighted evaluation for each pixel based on a boundary importance factor. The proposed wIoU can evaluate both contour and region by setting a boundary importance factor. We validated the effectiveness of wIoU on a dataset of 33 scenes and demonstrated its flexibility. Using the proposed metric, we expect more flexible and intuitive evaluation in semantic segmentation field are possible.
Transformers have been successful for many natural language processing tasks. However, applying transformers to the video domain for tasks such as long-term video generation and scene understanding has remained elusive due to the high computational complexity and the lack of natural tokenization. In this paper, we propose the Object-Centric Video Transformer (OCVT) which utilizes an object-centric approach for decomposing scenes into tokens suitable for use in a generative video transformer. By factoring the video into objects, our fully unsupervised model is able to learn complex spatio-temporal dynamics of multiple interacting objects in a scene and generate future frames of the video. Our model is also significantly more memory-efficient than pixel-based models and thus able to train on videos of length up to 70 frames with a single 48GB GPU. We compare our model with previous RNN-based approaches as well as other possible video transformer baselines. We demonstrate OCVT performs well when compared to baselines in generating future frames. OCVT also develops useful representations for video reasoning, achieving start-of-the-art performance on the CATER task.
Automotive Cyber-Physical Systems (ACPS) have attracted a significant amount of interest in the past few decades, while one of the most critical operations in these systems is the perception of the environment. Deep learning and, especially, the use of Deep Neural Networks (DNNs) provides impressive results in analyzing and understanding complex and dynamic scenes from visual data. The prediction horizons for those perception systems are very short and inference must often be performed in real time, stressing the need of transforming the original large pre-trained networks into new smaller models, by utilizing Model Compression and Acceleration (MCA) techniques. Our goal in this work is to investigate best practices for appropriately applying novel weight sharing techniques, optimizing the available variables and the training procedures towards the significant acceleration of widely adopted DNNs. Extensive evaluation studies carried out using various state-of-the-art DNN models in object detection and tracking experiments, provide details about the type of errors that manifest after the application of weight sharing techniques, resulting in significant acceleration gains with negligible accuracy losses.
We propose a novel dense mapping framework for sparse visual SLAM systems which leverages a compact scene representation. State-of-the-art sparse visual SLAM systems provide accurate and reliable estimates of the camera trajectory and locations of landmarks. While these sparse maps are useful for localization, they cannot be used for other tasks such as obstacle avoidance or scene understanding. In this paper we propose a dense mapping framework to complement sparse visual SLAM systems which takes as input the camera poses, keyframes and sparse points produced by the SLAM system and predicts a dense depth image for every keyframe. We build on CodeSLAM and use a variational autoencoder (VAE) which is conditioned on intensity, sparse depth and reprojection error images from sparse SLAM to predict an uncertainty-aware dense depth map. The use of a VAE then enables us to refine the dense depth images through multi-view optimization which improves the consistency of overlapping frames. Our mapper runs in a separate thread in parallel to the SLAM system in a loosely coupled manner. This flexible design allows for integration with arbitrary metric sparse SLAM systems without delaying the main SLAM process. Our dense mapper can be used not only for local mapping but also globally consistent dense 3D reconstruction through TSDF fusion. We demonstrate our system running with ORB-SLAM3 and show accurate dense depth estimation which could enable applications such as robotics and augmented reality.
Semantic segmentation for scene understanding is nowadays widely demanded, raising significant challenges for the algorithm efficiency, especially its applications on resource-limited platforms. Current segmentation models are trained and evaluated on massive high-resolution scene images ("data level") and suffer from the expensive computation arising from the required multi-scale aggregation("network level"). In both folds, the computational and energy costs in training and inference are notable due to the often desired large input resolutions and heavy computational burden of segmentation models. To this end, we propose DANCE, general automated DAta-Network Co-optimization for Efficient segmentation model training and inference. Distinct from existing efficient segmentation approaches that focus merely on light-weight network design, DANCE distinguishes itself as an automated simultaneous data-network co-optimization via both input data manipulation and network architecture slimming. Specifically, DANCE integrates automated data slimming which adaptively downsamples/drops input images and controls their corresponding contribution to the training loss guided by the images' spatial complexity. Such a downsampling operation, in addition to slimming down the cost associated with the input size directly, also shrinks the dynamic range of input object and context scales, therefore motivating us to also adaptively slim the network to match the downsampled data. Extensive experiments and ablating studies (on four SOTA segmentation models with three popular segmentation datasets under two training settings) demonstrate that DANCE can achieve "all-win" towards efficient segmentation(reduced training cost, less expensive inference, and better mean Intersection-over-Union (mIoU)).
Modeling complex phenomena typically involves the use of both discrete and continuous variables. Such a setting applies across a wide range of problems, from identifying trends in time-series data to performing effective compositional scene understanding in images. Here, we propose Hybrid Memoised Wake-Sleep (HMWS), an algorithm for effective inference in such hybrid discrete-continuous models. Prior approaches to learning suffer as they need to perform repeated expensive inner-loop discrete inference. We build on a recent approach, Memoised Wake-Sleep (MWS), which alleviates part of the problem by memoising discrete variables, and extend it to allow for a principled and effective way to handle continuous variables by learning a separate recognition model used for importance-sampling based approximate inference and marginalization. We evaluate HMWS in the GP-kernel learning and 3D scene understanding domains, and show that it outperforms current state-of-the-art inference methods.
Monocular depth estimation (MDE) is a fundamental task in many applications such as scene understanding and reconstruction. However, most of the existing methods rely on accurately labeled datasets. A weakly-supervised framework based on attention nested U-net (ANU) named as ANUW is introduced in this paper for cases with wrong labels. The ANUW is trained end-to-end to convert an input single RGB image into a depth image. It consists of a dense residual network structure, an adaptive weight channel attention (AWCA) module, a patch second non-local (PSNL) module and a soft label generation method. The dense residual network is the main body of the network to encode and decode the input. The AWCA module can adaptively adjust the channel weights to extract important features. The PSNL module implements the spatial attention mechanism through a second-order non-local method. The proposed soft label generation method uses the prior knowledge of the dataset to produce soft labels to replace false ones. The proposed ANUW is trained on a defective monocular depth dataset and the trained model is tested on three public datasets, and the results demonstrate the superiority of ANUW in comparison with the state-of-the-art MDE methods.
We present SynPick, a synthetic dataset for dynamic scene understanding in bin-picking scenarios. In contrast to existing datasets, our dataset is both situated in a realistic industrial application domain -- inspired by the well-known Amazon Robotics Challenge (ARC) -- and features dynamic scenes with authentic picking actions as chosen by our picking heuristic developed for the ARC 2017. The dataset is compatible with the popular BOP dataset format. We describe the dataset generation process in detail, including object arrangement generation and manipulation simulation using the NVIDIA PhysX physics engine. To cover a large action space, we perform untargeted and targeted picking actions, as well as random moving actions. To establish a baseline for object perception, a state-of-the-art pose estimation approach is evaluated on the dataset. We demonstrate the usefulness of tracking poses during manipulation instead of single-shot estimation even with a naive filtering approach. The generator source code and dataset are publicly available.
Cyber Physical Systems have been going into a transition phase from individual systems to a collecttives of systems that collaborate in order to achieve a highly complex cause, realizing a system of systems approach. The automotive domain has been making a transition to the system of system approach aiming to provide a series of emergent functionality like traffic management, collaborative car fleet management or large-scale automotive adaptation to physical environment thus providing significant environmental benefits (e.g air pollution reduction) and achieving significant societal impact. Similarly, large infrastructure domains, are evolving into global, highly integrated cyber-physical systems of systems covering all parts of the value chain. In practice, there are significant challenges in CPSoS applicability and usability to be addressed, i.e. even a small CPSoS such as a car consists several subsystems Decentralization of CPSoS appoints tasks to individual CPSs within the System of Systems. CPSoSs are heterogenous systems. They comprise of various, autonomous, CPSs, each one of them having unique performance capabilities, criticality level, priorities and pursued goals. all CPSs must also harmonically pursue system-based achievements and collaborate in order to make system-of-system based decisions and implement the CPSoS functionality. This survey will provide a comprehensive review on current best practices in connected cyberphysical systems. The basis of our investigation is a dual layer architecture encompassing a perception layer and a behavioral layer. Perception algorithms with respect to scene understanding (object detection and tracking, pose estimation), localization mapping and path planning are thoroughly investigated. Behavioural part focuses on decision making and human in the loop control.
As cameras are increasingly deployed in new application domains such as autonomous driving, performing 3D object detection on monocular images becomes an important task for visual scene understanding. Recent advances on monocular 3D object detection mainly rely on the ``pseudo-LiDAR'' generation, which performs monocular depth estimation and lifts the 2D pixels to pseudo 3D points. However, depth estimation from monocular images, due to its poor accuracy, leads to inevitable position shift of pseudo-LiDAR points within the object. Therefore, the predicted bounding boxes may suffer from inaccurate location and deformed shape. In this paper, we present a novel neighbor-voting method that incorporates neighbor predictions to ameliorate object detection from severely deformed pseudo-LiDAR point clouds. Specifically, each feature point around the object forms their own predictions, and then the ``consensus'' is achieved through voting. In this way, we can effectively combine the neighbors' predictions with local prediction and achieve more accurate 3D detection. To further enlarge the difference between the foreground region of interest (ROI) pseudo-LiDAR points and the background points, we also encode the ROI prediction scores of 2D foreground pixels into the corresponding pseudo-LiDAR points. We conduct extensive experiments on the KITTI benchmark to validate the merits of our proposed method. Our results on the bird's eye view detection outperform the state-of-the-art performance by a large margin, especially for the ``hard'' level detection.
Visual Commonsense Reasoning (VCR) predicts an answer with corresponding rationale, given a question-image input. VCR is a recently introduced visual scene understanding task with a wide range of applications, including visual question answering, automated vehicle systems, and clinical decision support. Previous approaches to solving the VCR task generally rely on pre-training or exploiting memory with long dependency relationship encoded models. However, these approaches suffer from a lack of generalizability and prior knowledge. In this paper we propose a dynamic working memory based cognitive VCR network, which stores accumulated commonsense between sentences to provide prior knowledge for inference. Extensive experiments show that the proposed model yields significant improvements over existing methods on the benchmark VCR dataset. Moreover, the proposed model provides intuitive interpretation into visual commonsense reasoning. A Python implementation of our mechanism is publicly available at https://github.com/tanjatang/DMVCR
Video segmentation -- partitioning video frames into multiple segments or objects -- plays a critical role in a broad range of practical applications, from enhancing visual effects in movie, to understanding scenes in autonomous driving, to creating virtual background in video conferencing. Recently, with the renaissance of connectionism in computer vision, there has been an influx of deep learning based approaches for video segmentation that have delivered compelling performance. In this survey, we comprehensively review two basic lines of research -- generic object segmentation (of unknown categories) in videos, and video semantic segmentation -- by introducing their respective task settings, background concepts, perceived need, development history, and main challenges. We also offer a detailed overview of representative literature on both methods and datasets. We further benchmark the reviewed methods on several well-known datasets. Finally, we point out open issues in this field, and suggest opportunities for further research. We also provide a public website to continuously track developments in this fast advancing field: https://github.com/tfzhou/VS-Survey.
Semantic segmentation is one of the basic, yet essential scene understanding tasks for an autonomous agent. The recent developments in supervised machine learning and neural networks have enjoyed great success in enhancing the performance of the state-of-the-art techniques for this task. However, their superior performance is highly reliant on the availability of a large-scale annotated dataset. In this paper, we propose a novel fully unsupervised semantic segmentation method, the so-called Information Maximization and Adversarial Regularization Segmentation (InMARS). Inspired by human perception which parses a scene into perceptual groups, rather than analyzing each pixel individually, our proposed approach first partitions an input image into meaningful regions (also known as superpixels). Next, it utilizes Mutual-Information-Maximization followed by an adversarial training strategy to cluster these regions into semantically meaningful classes. To customize an adversarial training scheme for the problem, we incorporate adversarial pixel noise along with spatial perturbations to impose photometrical and geometrical invariance on the deep neural network. Our experiments demonstrate that our method achieves the state-of-the-art performance on two commonly used unsupervised semantic segmentation datasets, COCO-Stuff, and Potsdam.
Camera-based passive dietary intake monitoring is able to continuously capture the eating episodes of a subject, recording rich visual information, such as the type and volume of food being consumed, as well as the eating behaviours of the subject. However, there currently is no method that is able to incorporate these visual clues and provide a comprehensive context of dietary intake from passive recording (e.g., is the subject sharing food with others, what food the subject is eating, and how much food is left in the bowl). On the other hand, privacy is a major concern while egocentric wearable cameras are used for capturing. In this paper, we propose a privacy-preserved secure solution (i.e., egocentric image captioning) for dietary assessment with passive monitoring, which unifies food recognition, volume estimation, and scene understanding. By converting images into rich text descriptions, nutritionists can assess individual dietary intake based on the captions instead of the original images, reducing the risk of privacy leakage from images. To this end, an egocentric dietary image captioning dataset has been built, which consists of in-the-wild images captured by head-worn and chest-worn cameras in field studies in Ghana. A novel transformer-based architecture is designed to caption egocentric dietary images. Comprehensive experiments have been conducted to evaluate the effectiveness and to justify the design of the proposed architecture for egocentric dietary image captioning. To the best of our knowledge, this is the first work that applies image captioning for dietary intake assessment in real life settings.
3D semantic scene completion and 2D semantic segmentation are two tightly correlated tasks that are both essential for indoor scene understanding, because they predict the same semantic classes, using positively correlated high-level features. Current methods use 2D features extracted from early-fused RGB-D images for 2D segmentation to improve 3D scene completion. We argue that this sequential scheme does not ensure these two tasks fully benefit each other, and present an Iterative Mutual Enhancement Network (IMENet) to solve them jointly, which interactively refines the two tasks at the late prediction stage. Specifically, two refinement modules are developed under a unified framework for the two tasks. The first is a 2D Deformable Context Pyramid (DCP) module, which receives the projection from the current 3D predictions to refine the 2D predictions. In turn, a 3D Deformable Depth Attention (DDA) module is proposed to leverage the reprojected results from 2D predictions to update the coarse 3D predictions. This iterative fusion happens to the stable high-level features of both tasks at a late stage. Extensive experiments on NYU and NYUCAD datasets verify the effectiveness of the proposed iterative late fusion scheme, and our approach outperforms the state of the art on both 3D semantic scene completion and 2D semantic segmentation.
Instance segmentation of images is an important tool for automated scene understanding. Neural networks are usually trained to optimize their overall performance in terms of accuracy. Meanwhile, in applications such as automated driving, an overlooked pedestrian seems more harmful than a falsely detected one. In this work, we present a false negative detection method for image sequences based on inconsistencies in time series of tracked instances given the availability of image sequences in online applications. As the number of instances can be greatly increased by this algorithm, we apply a false positive pruning using uncertainty estimates aggregated over instances. To this end, instance-wise metrics are constructed which characterize uncertainty and geometry of a given instance or are predicated on depth estimation. The proposed method serves as a post-processing step applicable to any neural network that can also be trained on single frames only. In our tests, we obtain an improved trade-off between false negative and false positive instances by our fused detection approach in comparison to the use of an ordinary score value provided by the instance segmentation network during inference.
Multiple human tracking is a fundamental problem for scene understanding. Although both accuracy and speed are required in real-world applications, recent tracking methods based on deep learning have focused on accuracy and require substantial running time. This study aims to improve running speed by performing human detection at a certain frame interval because it accounts for most of the running time. The question is how to maintain accuracy while skipping human detection. In this paper, we propose a method that complements the detection results with optical flow, based on the fact that someone's appearance does not change much between adjacent frames. To maintain the tracking accuracy, we introduce robust interest point selection within human regions and a tracking termination metric calculated by the distribution of the interest points. On the MOT20 dataset in the MOTChallenge, the proposed SDOF-Tracker achieved the best performance in terms of the total running speed while maintaining the MOTA metric. Our code is available at https://github.com/hitottiez/sdof-tracker.
We present OffRoadTranSeg, the first end-to-end framework for semi-supervised segmentation in unstructured outdoor environment using transformers and automatic data selection for labelling. The offroad segmentation is a scene understanding approach that is widely used in autonomous driving. The popular offroad segmentation method is to use fully connected convolution layers and large labelled data, however, due to class imbalance, there will be several mismatches and also some classes may not be detected. Our approach is to do the task of offroad segmentation in a semi-supervised manner. The aim is to provide a model where self supervised vision transformer is used to fine-tune offroad datasets with self-supervised data collection for labelling using depth estimation. The proposed method is validated on RELLIS-3D and RUGD offroad datasets. The experiments show that OffRoadTranSeg outperformed other state of the art models, and also solves the RELLIS-3D class imbalance problem.