We present Splatt3RFusion, a feed-forward neural network that, given a set of unposed and uncalibrated images, directly reconstructs a compact and high-quality 3D Gaussian Splat representation of a scene. Unlike prior pixel-aligned feed-forward methods that typically predict one 3D Gaussian primitive per pixel in each image - producing severe redundancy, duplication, and ghosting on one physical surface - our approach efficiently fuses points in 3D space through a multi-scale octree structure, yielding a compact and coherent representation. Built upon VGGT, a foundation model for pose-free 3D geometry prediction, Splatt3RFusion introduces a Gaussian prediction branch that infers primitive parameters using only photometric supervision. We also introduce the ability to control the number of 3D Gaussians generated at test-time, allowing for a controllable tradeoff between PSNR and the number of 3D Gaussian primitives used. The model is efficient, reducing both memory usage and rendering cost, while achieving state-of-the-art results on RealEstate10k and ScanNet++.

Mirror reflections are common in everyday environments and can provide stereo information within a single capture, as the real and reflected virtual views are visible simultaneously. We exploit this property by treating the reflection as an auxiliary view and designing a transformation that constructs a physically valid virtual camera, allowing direct pixel-domain generation of the virtual view while adhering to the real-world imaging process. This enables a multi-view stereo setup from a single image, simplifying the imaging process, making it compatible with powerful feedforward reconstruction models for generalizable and robust 3D reconstruction. To further exploit the geometric symmetry introduced by mirrors, we propose a symmetric-aware loss to refine pose estimation. Our framework also naturally extends to dynamic scenes, where each frame contains a mirror reflection, enabling efficient per-frame geometry recovery. For quantitative evaluation, we provide a fully customizable synthetic dataset of 16 Blender scenes, each with ground-truth point clouds and camera poses. Extensive experiments on real-world data and synthetic data are conducted to illustrate the effectiveness of our method. Code and Blender scenes are available in: https://github.com/jingwu2121/reflect3r.git.

Scene coordinate regression (SCR) has established itself as a promising learning-based approach to visual relocalization. After mere minutes of scene-specific training, SCR models estimate camera poses of query images with high accuracy. Still, SCR methods fall short of the generalization capabilities of more classical feature-matching approaches. When imaging conditions of query images, such as lighting or viewpoint, are too different from the training views, SCR models fail. Failing to generalize is an inherent limitation of previous SCR frameworks, since their training objective is to encode the training views in the weights of the coordinate regressor itself. The regressor essentially overfits to the training views, by design. We propose to separate the coordinate regressor and the map representation into a generic transformer and a scene-specific map code. This separation allows us to pre-train the transformer on tens of thousands of scenes. More importantly, it allows us to train the transformer to generalize from mapping images to unseen query images during pre-training. We demonstrate on multiple challenging relocalization datasets that our method, ACE-G, leads to significantly increased robustness while keeping the computational footprint attractive

We leverage 3D Gaussian Splatting (3DGS) as a scene representation and propose a novel test-time camera pose refinement (CPR) framework, GS-CPR. This framework enhances the localization accuracy of state-of-the-art absolute pose regression and scene coordinate regression methods. The 3DGS model renders high-quality synthetic images and depth maps to facilitate the establishment of 2D-3D correspondences. GS-CPR obviates the need for training feature extractors or descriptors by operating directly on RGB images, utilizing the 3D foundation model, MASt3R, for precise 2D matching. To improve the robustness of our model in challenging outdoor environments, we incorporate an exposure-adaptive module within the 3DGS framework. Consequently, GS-CPR enables efficient one-shot pose refinement given a single RGB query and a coarse initial pose estimation. Our proposed approach surpasses leading NeRF-based optimization methods in both accuracy and runtime across indoor and outdoor visual localization benchmarks, achieving new state-of-the-art accuracy on two indoor datasets.

We introduce Oxford Day-and-Night, a large-scale, egocentric dataset for novel view synthesis (NVS) and visual relocalisation under challenging lighting conditions. Existing datasets often lack crucial combinations of features such as ground-truth 3D geometry, wide-ranging lighting variation, and full 6DoF motion. Oxford Day-and-Night addresses these gaps by leveraging Meta ARIA glasses to capture egocentric video and applying multi-session SLAM to estimate camera poses, reconstruct 3D point clouds, and align sequences captured under varying lighting conditions, including both day and night. The dataset spans over 30 km of recorded trajectories and covers an area of $40{,}000\mathrm{m}^2$, offering a rich foundation for egocentric 3D vision research. It supports two core benchmarks, NVS and relocalisation, providing a unique platform for evaluating models in realistic and diverse environments. Project page: https://oxdan.active.vision/

Given two images, we can estimate the relative camera pose between them by establishing image-to-image correspondences. Usually, correspondences are 2D-to-2D and the pose we estimate is defined only up to scale. Some applications, aiming at instant augmented reality anywhere, require scale-metric pose estimates, and hence, they rely on external depth estimators to recover the scale. We present MicKey, a keypoint matching pipeline that is able to predict metric correspondences in 3D camera space. By learning to match 3D coordinates across images, we are able to infer the metric relative pose without depth measurements. Depth measurements are also not required for training, nor are scene reconstructions or image overlap information. MicKey is supervised only by pairs of images and their relative poses. MicKey achieves state-of-the-art performance on the Map-Free Relocalisation benchmark while requiring less supervision than competing approaches.

Absolute Pose Regression (APR) methods use deep neural networks to directly regress camera poses from RGB images. However, the predominant APR architectures only rely on 2D operations during inference, resulting in limited accuracy of pose estimation due to the lack of 3D geometry constraints or priors. In this work, we propose a test-time refinement pipeline that leverages implicit geometric constraints using a robust feature field to enhance the ability of APR methods to use 3D information during inference. We also introduce a novel Neural Feature Synthesizer (NeFeS) model, which encodes 3D geometric features during training and directly renders dense novel view features at test time to refine APR methods. To enhance the robustness of our model, we introduce a feature fusion module and a progressive training strategy. Our proposed method achieves state-of-the-art single-image APR accuracy on indoor and outdoor datasets. Code will be released at https://github.com/ActiveVisionLab/NeFeS.

Pose regression networks predict the camera pose of a query image relative to a known environment. Within this family of methods, absolute pose regression (APR) has recently shown promising accuracy in the range of a few centimeters in position error. APR networks encode the scene geometry implicitly in their weights. To achieve high accuracy, they require vast amounts of training data that, realistically, can only be created using novel view synthesis in a days-long process. This process has to be repeated for each new scene again and again. We present a new approach to pose regression, map-relative pose regression (marepo), that satisfies the data hunger of the pose regression network in a scene-agnostic fashion. We condition the pose regressor on a scene-specific map representation such that its pose predictions are relative to the scene map. This allows us to train the pose regressor across hundreds of scenes to learn the generic relation between a scene-specific map representation and the camera pose. Our map-relative pose regressor can be applied to new map representations immediately or after mere minutes of fine-tuning for the highest accuracy. Our approach outperforms previous pose regression methods by far on two public datasets, indoor and outdoor. Code is available: https://nianticlabs.github.io/marepo.

In this paper, we address the challenge of matching semantically similar keypoints across image pairs. Existing research indicates that the intermediate output of the UNet within the Stable Diffusion (SD) can serve as robust image feature maps for such a matching task. We demonstrate that by employing a basic prompt tuning technique, the inherent potential of Stable Diffusion can be harnessed, resulting in a significant enhancement in accuracy over previous approaches. We further introduce a novel conditional prompting module that conditions the prompt on the local details of the input image pairs, leading to a further improvement in performance. We designate our approach as SD4Match, short for Stable Diffusion for Semantic Matching. Comprehensive evaluations of SD4Match on the PF-Pascal, PF-Willow, and SPair-71k datasets show that it sets new benchmarks in accuracy across all these datasets. Particularly, SD4Match outperforms the previous state-of-the-art by a margin of 12 percentage points on the challenging SPair-71k dataset. Code is available at the project website: https://sd4match.active.vision/.

Neural surface reconstruction is sensitive to the camera pose noise, even when state-of-the-art pose estimators like COLMAP or ARKit are used. Existing Pose-NeRF joint optimisation methods have struggled to improve pose accuracy in challenging real-world scenarios. To overcome the challenges, we introduce the pose residual field (PoRF), a novel implicit representation that uses an MLP for regressing pose updates. Compared with the conventional per-frame pose parameter optimisation, this new representation is more robust due to parameter sharing that leverages global information over the entire sequence. Furthermore, we propose an epipolar geometry loss to enhance the supervision that leverages the correspondences exported from COLMAP results without the extra computational overhead. Our method yields promising results. On the DTU dataset, we reduce the rotation error of COLMAP poses by 78\%, leading to the reduced reconstruction Chamfer distance from 3.48mm to 0.85mm. On the MobileBrick dataset that contains casually captured unbounded 360-degree videos, our method refines ARKit poses and improves the reconstruction F1 score from 69.18 to 75.67, outperforming that with the provided ground-truth pose (75.14). These achievements demonstrate the efficacy of our approach in refining camera poses and improving the accuracy of neural surface reconstruction in real-world scenarios.

Absolute Pose Regressors (APRs) directly estimate camera poses from monocular images, but their accuracy is unstable for different queries. Uncertainty-aware APRs provide uncertainty information on the estimated pose, alleviating the impact of these unreliable predictions. However, existing uncertainty modelling techniques are often coupled with a specific APR architecture, resulting in suboptimal performance compared to state-of-the-art (SOTA) APR methods. This work introduces a novel APR-agnostic framework, HR-APR, that formulates uncertainty estimation as cosine similarity estimation between the query and database features. It does not rely on or affect APR network architecture, which is flexible and computationally efficient. In addition, we take advantage of the uncertainty for pose refinement to enhance the performance of APR. The extensive experiments demonstrate the effectiveness of our framework, reducing 27.4% and 15.2% of computational overhead on the 7Scenes and Cambridge Landmarks datasets while maintaining the SOTA accuracy in single-image APRs.

We address the problem of establishing accurate correspondences between two images. We present a flexible framework that can easily adapt to both geometric and semantic matching. Our contribution consists of three parts. Firstly, we propose an end-to-end trainable framework that uses the coarse-to-fine matching strategy to accurately find the correspondences. We generate feature maps in two levels of resolution, enforce the neighbourhood consensus constraint on the coarse feature maps by 4D convolutions and use the resulting correlation map to regulate the matches from the fine feature maps. Secondly, we present three variants of the model with different focuses. Namely, a universal correspondence model named DualRC that is suitable for both geometric and semantic matching, an efficient model named DualRC-L tailored for geometric matching with a lightweight neighbourhood consensus module that significantly accelerates the pipeline for high-resolution input images, and the DualRC-D model in which we propose a novel dynamically adaptive neighbourhood consensus module (DyANC) that dynamically selects the most suitable non-isotropic 4D convolutional kernels with the proper neighbourhood size to account for the scale variation. Last, we thoroughly experiment on public benchmarks for both geometric and semantic matching, showing superior performance in both cases.

Camera pose estimation for two-view geometry traditionally relies on RANSAC. Normally, a multitude of image correspondences leads to a pool of proposed hypotheses, which are then scored to find a winning model. The inlier count is generally regarded as a reliable indicator of “consensus”. We examine this scoring heuristic, and find that it favors disappointing models under certain circumstances. As a remedy, we propose the Fundamental Scoring Network (FSNet), which infers a score for a pair of overlap-ping images and any proposed fundamental matrix. It does not rely on sparse correspondences, but rather embodies a two-view geometry model through an epipolar attention mechanism that predicts the pose error of the two images. FSNet can be incorporated into traditional RANSAC loops. We evaluate FSNet onfundamental and essential matrix estimation on indoor and outdoor datasets, and establish that FSNet can successfully identify good poses for pairs of images with few or unreliable correspondences. Besides, we show that naively combining FSNet with MAGSAC++ scoring approach achieves state of the art results.

Learning-based visual relocalizers exhibit leading pose accuracy, but require hours or days of training. Since training needs to happen on each new scene again, long training times make learning-based relocalization impractical for most applications, despite its promise of high accuracy. In this paper we show how such a system can actually achieve the same accuracy in less than 5 minutes. We start from the obvious: a relocalization network can be split in a scene-agnostic feature backbone, and a scene-specific prediction head. Less obvious: using an MLP prediction head allows us to optimize across thousands of view points simultaneously in each single training iteration. This leads to stable and extremely fast convergence. Furthermore, we substitute effective but slow end-to-end training using a robust pose solver with a curriculum over a reprojection loss. Our approach does not require privileged knowledge, such a depth maps or a 3D model, for speedy training. Overall, our approach is up to 300x faster in mapping than state-of-the-art scene coordinate regression, while keeping accuracy on par. Code is available: https://nianticlabs.github.io/ace

High-quality 3D ground-truth shapes are critical for 3D object reconstruction evaluation. However, it is difficult to create a replica of an object in reality, and even 3D reconstructions generated by 3D scanners have artefacts that cause biases in evaluation. To address this issue, we introduce a novel multi-view RGBD dataset captured using a mobile device, which includes highly precise 3D ground-truth annotations for 153 object models featuring a diverse set of 3D structures. We obtain precise 3D ground-truth shape without relying on high-end 3D scanners by utilising LEGO models with known geometry as the 3D structures for image capture. The distinct data modality offered by high-resolution RGB images and low-resolution depth maps captured on a mobile device, when combined with precise 3D geometry annotations, presents a unique opportunity for future research on high-fidelity 3D reconstruction. Furthermore, we evaluate a range of 3D reconstruction algorithms on the proposed dataset.

Prior quantization methods focus on producing networks for fast and lightweight inference. However, the cost of unquantised training is overlooked, despite requiring significantly more time and energy than inference. We present a method for quantizing convolutional neural networks for efficient training. Quantizing gradients is challenging because it requires higher granularity and their values span a wider range than the weight and feature maps. We propose an extension of the Channel-wise Block Floating Point format that allows for quick gradient computation, using a minimal amount of quantization time. This is achieved through sharing an exponent across both depth and batch dimensions in order to quantize tensors once and reuse them during backpropagation. We test our method using standard models such as AlexNet, VGG, and ResNet, on the CIFAR10, SVHN and ImageNet datasets. We show no loss of accuracy when quantizing AlexNet weights, activations and gradients to only 4 bits training ImageNet.

The prowess that makes few-shot learning desirable in medical image analysis is the efficient use of the support image data, which are labelled to classify or segment new classes, a task that otherwise requires substantially more training images and expert annotations. This work describes a fully 3D prototypical few-shot segmentation algorithm, such that the trained networks can be effectively adapted to clinically interesting structures that are absent in training, using only a few labelled images from a different institute. First, to compensate for the widely recognised spatial variability between institutions in episodic adaptation of novel classes, a novel spatial registration mechanism is integrated into prototypical learning, consisting of a segmentation head and an spatial alignment module. Second, to assist the training with observed imperfect alignment, support mask conditioning module is proposed to further utilise the annotation available from the support images. Extensive experiments are presented in an application of segmenting eight anatomical structures important for interventional planning, using a data set of 589 pelvic T2-weighted MR images, acquired at seven institutes. The results demonstrate the efficacy in each of the 3D formulation, the spatial registration, and the support mask conditioning, all of which made positive contributions independently or collectively. Compared with the previously proposed 2D alternatives, the few-shot segmentation performance was improved with statistical significance, regardless whether the support data come from the same or different institutes.

Dense 3D reconstruction from a stream of depth images is the key to many mixed reality and robotic applications. Although methods based on Truncated Signed Distance Function (TSDF) Fusion have advanced the field over the years, the TSDF volume representation is confronted with striking a balance between the robustness to noisy measurements and maintaining the level of detail. We present Bi-level Neural Volume Fusion (BNV-Fusion), which leverages recent advances in neural implicit representations and neural rendering for dense 3D reconstruction. In order to incrementally integrate new depth maps into a global neural implicit representation, we propose a novel bi-level fusion strategy that considers both efficiency and reconstruction quality by design. We evaluate the proposed method on multiple datasets quantitatively and qualitatively, demonstrating a significant improvement over existing methods.

We present a relocalization pipeline, which combines an absolute pose regression (APR) network with a novel view synthesis based direct matching module, offering superior accuracy while maintaining low inference time. Our contribution is twofold: i) we design a direct matching module that supplies a photometric supervision signal to refine the pose regression network via differentiable rendering; ii) we show that our method can easily cope with additional unlabeled data without the need for external supervision such as traditional visual odometry or pose graph optimization. As a result, our method achieves state-of-the-art performance among all other single-image APR methods on the 7-Scenes benchmark and the LLFF dataset.

Self-supervised monocular depth estimation networks are trained to predict scene depth using nearby frames as a supervision signal during training. However, for many applications, sequence information in the form of video frames is also available at test time. The vast majority of monocular networks do not make use of this extra signal, thus ignoring valuable information that could be used to improve the predicted depth. Those that do, either use computationally expensive test-time refinement techniques or off-the- shelf recurrent networks, which only indirectly make use of the geometric information that is inherently available.We propose ManyDepth, an adaptive approach to dense depth estimation that can make use of sequence information at test time, when it is available. Taking inspiration from multi-view stereo, we propose a deep end-to-end cost volume based approach that is trained using self-supervision only. We present a novel consistency loss that encourages the network to ignore the cost volume when it is deemed unreliable, e.g. in the case of moving objects, and an augmentation scheme to cope with static cameras. Our detailed experiments on both KITTI and Cityscapes show that we outperform all published self-supervised baselines, including those that use single or multiple frames at test time.

We present LaLaLoc to localise in environments without the need for prior visitation, and in a manner that is robust to large changes in scene appearance, such as a full rearrangement of furniture. Specifically, LaLaLoc performs localisation through latent representations of room layout. LaLaLoc learns a rich embedding space shared between RGB panoramas and layouts inferred from a known floor plan that encodes the structural similarity between locations. Further, LaLaLoc introduces direct, cross-modal pose optimisation in its latent space. Thus, LaLaLoc enables fine-grained pose estimation in a scene without the need for prior visitation, as well as being robust to dynamics, such as a change in furniture configuration. We show that in a domestic environment LaLaLoc is able to accurately localise a single RGB panorama image to within 8.3cm, given only a floor plan as a prior.

Aggregating features from different depths of a network is widely adopted to improve the network capability. Lots of modern architectures are equipped with skip connections, which actually makes the feature aggregation happen in all these networks. Since different features tell different semantic meanings, there are inconsistencies and incompatibilities to be solved. However, existing works naively blend deep features via element-wise summation or concatenation with a convolution behind. Better feature aggregation method beyond summation or concatenation is rarely explored. In this paper, given two layers of features to be aggregated together, we first detect and identify where and what needs to be updated in one layer, then replace the feature at the identified location with the information of the other layer This process, which we call DEtect-rePLAce (DEPLA), enables us to avoid inconsistent patterns while keeping useful information in the merged outputs. Experimental results demonstrate our method largely boosts multiple baselines e.g. ResNet, FishNet and FPN on three major vision tasks including ImageNet classification, MS COCO object detection and instance segmentation.

Full-motion cost volumes play a central role in current state-of-the-art optical flow methods. However, constructed using simple feature correlations, they lack the ability to encapsulate prior, or even non-local knowledge. This creates artifacts in poorly constrained ambiguous regions, such as occluded and textureless areas. We propose a separable cost volume module, a drop-in replacement to correlation cost volumes, that uses non-local aggregation layers to exploit global context cues and prior knowledge, in order to disambiguate motions in these regions. Our method leads both the now standard Sintel and KITTI optical flow benchmarks in terms of accuracy, and is also shown to generalize better from synthetic to real data.

In this paper, we tackle the task of establishing dense visual correspondences between images containing objects of the same category. This is a challenging task due to large intra-class variations and a lack of dense pixel level annotations. We propose a convolutional neural network architecture, called adaptive neighbourhood consensus network (ANC-Net), that can be trained end-to-end with sparse key-point annotations, to handle this challenge. At the core of ANC-Net is our proposed non-isotropic 4D convolution kernel, which forms the building block for the adaptive neighbourhood consensus module for robust matching. We also introduce a simple and efficient multi-scale self-similarity module in ANC-Net to make the learned feature robust to intra-class variations. Furthermore, we propose a novel orthogonal loss that can enforce the one-to-one matching constraint. We thoroughly evaluate the effectiveness of our method on various benchmarks, where it substantially outperforms state-of-the-art methods.

We propose a robust baseline method for instance segmentation which are specially designed for large-scale outdoor LiDAR point clouds. Our method includes a novel dense feature encoding technique, allowing the localization and segmentation of small, far-away objects, a simple but effective solution for single-shot instance prediction and effective strategies for handling severe class imbalances. Since there is no public dataset for the study of LiDAR instance segmentation, we also build a new publicly available LiDAR point cloud dataset to include both precise 3D bounding box and point-wise labels for instance segmentation, while still being about 3~20 times as large as other existing LiDAR datasets. The dataset will be published at https://github.com/feihuzhang/LiDARSeg.

We tackle the problem of establishing dense pixel-wise correspondences between a pair of images. In this work, we introduce Dual-Resolution Correspondence Networks (DualRC-Net), to obtain pixel-wise correspondences in a coarse-to-fine manner. DualRC-Net extracts both coarse- and fine- resolution feature maps. The coarse maps are used to produce a full but coarse 4D correlation tensor, which is then refined by a learnable neighbourhood consensus module. The fine-resolution feature maps are used to obtain the final dense correspondences guided by the refined coarse 4D correlation tensor. The selected coarse-resolution matching scores allow the fine-resolution features to focus only on a limited number of possible matches with high confidence. In this way, DualRC-Net dramatically increases matching reliability and localisation accuracy, while avoiding to apply the expensive 4D convolution kernels on fine-resolution feature maps. We comprehensively evaluate our method on large-scale public benchmarks including HPatches, InLoc, and Aachen Day-Night. It achieves state-of-the-art results on all of them.

We present FlowNet3D++, a deep scene flow estimation network. Inspired by classical methods, FlowNet3D++ incorporates geometric constraints in the form of point-toplane distance and angular alignment between individual vectors in the flow field, into FlowNet3D [21]. We demonstrate that the addition of these geometric loss terms improves the previous state-of-art FlowNet3D accuracy from 57.85% to 63.43%. To further demonstrate the effectiveness of our geometric constraints, we propose a benchmark for flow estimation on the task of dynamic 3D reconstruction, thus providing a more holistic and practical measure of performance than the breakdown of individual metrics previously used to evaluate scene flow. This is made possible through the contribution of a novel pipeline to integrate point-based scene flow predictions into a global dense volume. FlowNet3D++ achieves up to a 15.0% reduction in reconstruction error over FlowNet3D, and up to a 35.2% improvement over KillingFusion [32] alone. We will release our scene flow estimation code later.

Camera pose estimation is an important problem in computer vision, with applications as diverse as simultaneous localisation and mapping, virtual/augmented reality and navigation. Common techniques match the current image against keyframes with known poses coming from a tracker, directly regress the pose, or establish correspondences between keypoints in the current image and points in the scene in order to estimate the pose. In recent years, regression forests have become a popular alternative to establish such correspondences. They achieve accurate results, but have traditionally needed to be trained offline on the target scene, preventing relocalisation in new environments. Recently, we showed how to circumvent this limitation by adapting a pre-trained forest to a new scene on the fly. The adapted forests achieved relocalisation performance that was on par with that of offline forests, and our approach was able to estimate the camera pose in close to real time, which made it desirable for systems that require online relocalisation. In this paper, we present an extension of this work that achieves significantly better relocalisation performance whilst running fully in real time. To achieve this, we make several changes to the original approach: (i) instead of simply accepting the camera pose hypothesis produced by RANSAC without question, we make it possible to score the final few hypotheses it considers using a geometric approach and select the most promising one; (ii) we chain several instantiations of our relocaliser (with different parameter settings) together in a cascade, allowing us to try faster but less accurate relocalisation first, only falling back to slower, more accurate relocalisation as necessary; and (iii) we tune the parameters of our cascade, and the individual relocalisers it contains, to achieve effective overall performance. Taken together, these changes allow us to significantly improve upon the performance our original state-of-the-art method was able to achieve on the well-known 7-Scenes and Stanford 4 Scenes benchmarks. As additional contributions, we present a novel way of visualising the internal behaviour of our forests, and use the insights gleaned from this to show how to entirely circumvent the need to pre-train a forest on a generic scene.

In the stereo matching task, matching cost aggregation is crucial in both traditional methods and deep neural network models in order to accurately estimate disparities. We propose two novel neural net layers, aimed at capturing local and the whole-image cost dependencies respectively. The first is a semi-global aggregation layer which is a differentiable approximation of the semi-global matching, the second is the local guided aggregation layer which follows a traditional cost filtering strategy to refine thin structures. These two layers can be used to replace the widely used 3D convolutional layer which is computationally costly and memory-consuming as it has cubic computational/memory complexity. In the experiments, we show that nets with a two-layer guided aggregation block easily outperform the state-of-the-art GC-Net which has nineteen 3D convolutional layers. We also train a deep guided aggregation network (GA-Net) which gets better accuracies than state-of-the-art methods on both Scene Flow dataset and KITTI benchmarks.

Quantization is a popular way of increasing the speed and lowering the memory usage of Convolution Neural Networks (CNNs). When labelled training data is available, network weights and activations have successfully been quantized down to 1-bit. The same cannot be said about the scenario when labelled training data is not available, e.g. when quantizing a pre-trained model, where current approaches show, at best, no loss of accuracy at 8-bit quantizations. We introduce DSConv, a flexible quantized convolution operator that replaces single-precision operations with their far less expensive integer counterparts, while maintaining the probability distributions over both the kernel weights and the outputs. We test our model as a plug-and-play replacement for standard convolution on most popular neural network architectures, ResNet, DenseNet, GoogLeNet, AlexNet and VGG-Net and demonstrate state-of-the-art results, with less than 1% loss of accuracy, without retraining, using only 4-bit quantization. We also show how a distillation-based adaptation stage with unlabelled data can improve results even further.

We propose a novel end-to-end trainable, deep, encoder-decoder architecture for single-pass semantic segmentation. Our approach is based on a cascaded architecture with feature-level long-range skip connections. The encoder incorporates the structure of ResNeXt's residual building blocks and adopts the strategy of repeating a building block that aggregates a set of transformations with the same topology. The decoder features a novel architecture, consisting of blocks, that (i) capture context information, (ii) generate semantic features, and (iii) enable fusion between different output resolutions. Crucially, we introduce dense decoder shortcut connections to allow decoder blocks to use semantic feature maps from all previous decoder levels, i.e. from all higher-level feature maps. The dense decoder connections allow for effective information propagation from one decoder block to another, as well as for multi-level feature fusion that significantly improves the accuracy. Importantly, these connections allow our method to obtain state-of-the-art performance on several challenging datasets, without the need of time-consuming multi-scale averaging of previous works.

Modern visual odometry algorithms utilize sparse point-based features for tracking due to their low computational cost. Current state-of-the-art methods are split between indirect methods that process features extracted from the image, and indirect methods that deal directly on pixel intensities. In recent years, line-based features have been used in SLAM and have shown an increase in performance albeit with an increase in computational cost. In this paper, we propose an extension to a point-based direct monocular visual odometry method. Here we that uses lines to guide keypoint selection rather than acting as features. Points on a line are treated as stronger keypoints than those in other parts of the image, steering point-selection away from less distinctive points and thereby increasing efficiency. By combining intensity and geometry information from a set of points on a line, accuracy may also be increased.

The model produced by 3D reconstruction algorithm is usually represented by voxels. The management of these voxels is usually divided into two categories: ordered and unordered methods. The ordered method holds too many empty voxels to maintain data order which leads to a low storage efficiency. On the contrary, the unordered method keeps massive index data to only store nonempty voxels. In this paper, we design a new data management method for real-time indoor 3D reconstruction, called Structured Skip List (SSL). The SSL can be treated as a semi-ordered method, because the advantages of both the ordered and unordered methods are taken into account: 1) it only holds nonempty voxels similar to the unordered method; 2) the structured information is introduced to reduce the storage space of index data. By these designs, the SSL has a better performance on storage efficiency. To handle the data collision in voxel allocation, a hash allocation list (HAL) is proposed. The length of each Skip List is kept balanced by fusing the IMU (Inertial Measurement Unit) information for a high operation efficiency. The storage efficiency analysis of different data management methods is shown in this paper. What's more, exhaustive investigation is carried out on several datasets with these methods. The experimental result demonstrates that our design can achieve a high storage efficiency with little time loss compared to the state-of-the-art methods.

Without knowledge of the absolute baseline between images, the scale of a map from a single-camera simultaneous localization and mapping system is subject to calamitous drift over time. We describe a monocular approach that in addition to point measurements also considers object detections to resolve this scale ambiguity and drift. By placing an expectation on the size of the objects, the scale estimation can be seamlessly integrated into a bundle adjustment. When object observations are available, the local scale of the map is then determined jointly with the camera pose in local adjustments. Unlike many previous visual odometry methods, our approach does not impose restrictions such as constant camera height or planar roadways, and is therefore more widely applicable. We evaluate our approach on the KITTI data set and show that it reduces scale drift over long-range outdoor sequences with a total length of 40 km. As the scale of objects is known absolutely, metric accuracy is obtained for all sequences. Qualitative evaluation is also performed on video footage from a hand-held camera.

Reconstructing dense, volumetric models of real-world 3D scenes is important for many tasks, but capturing large scenes can take significant time, and the risk of transient changes to the scene goes up as the capture time increases. These are good reasons to want instead to capture several smaller sub-scenes that can be joined to make the whole scene. Achieving this has traditionally been difficult: joining sub-scenes that may never have been viewed from the same angle requires a high-quality camera relocaliser that can cope with novel poses, and tracking drift in each sub-scene can prevent them from being joined to make a consistent overall scene. Recent advances, however, have significantly improved our ability to capture medium-sized sub-scenes with little to no tracking drift: real-time globally consistent reconstruction systems can close loops and re-integrate the scene surface on the fly, whilst new visual-inertial odometry approaches can significantly reduce tracking drift during live reconstruction. Moreover, high-quality regression forest-based relocalisers have recently been made more practical by the introduction of a method to allow them to be trained and used online. In this paper, we leverage these advances to present what to our knowledge is the first system to allow multiple users to collaborate interactively to reconstruct dense, voxel-based models of whole buildings using only consumer-grade hardware, a task that has traditionally been both time-consuming and dependent on the availability of specialised hardware. Using our system, an entire house or lab can be reconstructed in under half an hour and at a far lower cost than was previously possible.

Monocular visual localization and mapping algorithms are able to estimate the environment only up to scale, a degree of freedom which leads to scale drift, difficulty closing loops, and eventual failure. This paper describes an image-driven approach for scale-drift correction which uses a convolutional neural network to infer the speed of the camera from successive monocular video frames. We obtain continuous drift correction, avoiding the need for explicit higher-level representations of the map to resolve scale. We also propose a novel method of including speed estimates as a regularizer in bundle adjustment which avoids the pitfalls of sudden imposition of scale knowledge. We demonstrate our approach using long-distance sequences for which ground truth is available, and find output that is essentially free of scale drift. We compare the performance with number of other methods for scale-drift correction from monocular data, and show that our solution achieves more accurate results.

In recent years, major advances have been made in 3D scene reconstruction. However, much less progress has been made for objects, which can exhibit far fewer unambiguous geometric/texture cues than a full scene, and thus are much harder to track against. In this work we present a novel probabilistic object reconstruction framework that simultaneously allows for online, implicit deformation of the objects surface to reduce tracking drift and handle loop closure events. Coupled with our probabilistic formulation is the use of a multi subsegment representation of the object, used to enforce global consistency, with segmentation of the object built in to the formulation. Finally, we employ a CRF framework to refine the overall segmentation, defined by a probability field over the object. We present compelling results over the current state-of-the-art object reconstruction work and demonstrate robustness and consistency w.r.t. established dense SLAM frameworks.

We describe a novel probabilistic framework for real-time tracking of multiple objects from combined depth-colour imagery. Object shape is represented implicitly using 3D signed distance functions. Probabilistic generative models based on these functions are developed to account for the observed RGB-D imagery, and tracking is posed as a maximum a posteriori problem. We present first a method suited to tracking a single rigid 3D object, and then generalise this to multiple objects by combining distance functions into a shape union in the frame of the camera. This second model accounts for similarity and proximity between objects, and leads to robust real-time tracking without recourse to bolt-on or ad-hoc collision detection.

Our abilities in scene understanding, which allow us to perceive the 3D structure of our surroundings and intuitively recognise the objects we see, are things that we largely take for granted, but for robots, the task of understanding large scenes quickly remains extremely challenging. Recently, scene understanding approaches based on 3D reconstruction and semantic segmentation have become popular, but existing methods either do not scale, fail outdoors, provide only sparse reconstructions or are rather slow. In this paper, we build on a recent hash-based technique for large-scale fusion and an efficient mean-field inference algorithm for densely-connected CRFs to present what to our knowledge is the first system that can perform dense, large-scale, outdoor semantic reconstruction of a scene in (near) real time. We also present a `semantic fusion' approach that allows us to handle dynamic objects more effectively than previous approaches. We demonstrate the effectiveness of our approach on the KITTI dataset, and provide qualitative and quantitative results showing high-quality dense reconstruction and labelling of a number of scenes.

Figure-ground segmentation from bounding box input, provided either automatically or manually, has been extremely popular in the last decade and influenced various applications. A lot of research has focused on high-quality segmentation, using complex formulations which often lead to slow techniques, and often hamper practical usage. In this paper we demonstrate a very fast segmentation technique which still achieves very high quality results. We propose to replace the time consuming iterative refinement of global colour models in traditional GrabCut formulation by a densely connected crf. To motivate this decision, we show that a dense crf implicitly models unnormalized global colour models for foreground and background. Such relationship provides insightful analysis to bridge between dense crf and GrabCut functional. We extensively evaluate our algorithm using two famous benchmarks. Our experimental results demonstrated that the proposed algorithm achieves an order of magnitude (10×) speed-up with respect to the closest competitor, and at the same time achieves a considerably higher accuracy.

Volumetric methods provide efficient, flexible and simple ways of integrating multiple depth images into a full 3D model. They provide dense and photorealistic 3D reconstructions, and parallelised implementations on GPUs achieve real-time performance on modern graphics hardware. To run such methods on mobile devices, providing users with freedom of movement and instantaneous reconstruction feedback, remains challenging however. In this paper we present a range of modifications to existing volumetric integration methods based on voxel block hashing, considerably improving their performance and making them applicable to tablet computer applications. We present (i) optimisations for the basic data structure, and its allocation and integration; (ii) a highly optimised raycasting pipeline; and (iii) extensions to the camera tracker to incorporate IMU data. In total, our system thus achieves frame rates up 47 Hz on a Nvidia Shield Tablet and 910 Hz on a Nvidia GTX Titan XGPU, or even beyond 1.1 kHz without visualisation.

We present a novel framework for jointly tracking a camera in 3D and reconstructing the 3D model of an observed object. Due to the region based approach, our formulation can handle untextured objects, partial occlusions, motion blur, dynamic backgrounds and imperfect lighting. Our formulation also allows for a very efficient implementation which achieves real-time performance on a mobile phone, by running the pose estimation and the shape optimisation in parallel. We use a level set based pose estimation but completely avoid the, typically required, explicit computation of a global distance. This leads to tracking rates of more than 100 Hz on a desktop PC and 30 Hz on a mobile phone. Further, we incorporate additional orientation information from the phone's inertial sensor which helps us resolve the tracking ambiguities inherent to region based formulations. The reconstruction step first probabilistically integrates 2D image statistics from selected keyframes into a 3D volume, and then imposes coherency and compactness using a total variational regularisation term. The global optimum of the overall energy function is found using a continuous max-flow algorithm and we show that, similar to tracking, the integration of per voxel posteriors instead of likelihoods improves the precision and accuracy of the reconstruction.

Most current approaches for 3D object tracking rely on distinctive object appearances. While several such trackers can be instantiated to track multiple objects independently, this not only neglects that objects should not occupy the same space in 3D, but also fails when objects have highly similar or identical appearances. In this paper we develop a probabilistic graphical model that accounts for similarity and proximity and leads to robust real-time tracking of multiple objects from RGB-D data, without recourse to bolton collision detection.

We propose a formulation of monocular SLAM which combines live dense reconstruction with shape priors-based 3D tracking and reconstruction. Current live dense SLAM approaches are limited to the reconstruction of visible surfaces. Moreover, most of them are based on the minimisation of a photo-consistency error, which usually makes them sensitive to specularities. In the 3D pose recovery literature, problems caused by imperfect and ambiguous image information have been dealt with by using prior shape knowledge. At the same time, the success of depth sensors has shown that combining joint image and depth information drastically increases the robustness of the classical monocular 3D tracking and 3D reconstruction approaches. In this work we link dense SLAM to 3D object pose and shape recovery. More specifically, we automatically augment our SLAM system with object specific identity, together with 6D pose and additional shape degrees of freedom for the object(s) of known class in the scene, combining image data and depth information for the pose and shape recovery. This leads to a system that allows for full scaled 3D reconstruction with the known object(s) segmented from the scene. The segmentation enhances the clarity, accuracy and completeness of the maps built by the dense SLAM system, while the dense 3D data aids the segmentation process, yielding faster and more reliable convergence than when using 2D image data alone.

We introduce a probabilistic framework for simultaneous tracking and reconstruction of 3D rigid objects using an RGB-D camera. The tracking problem is handled using a bag-of-pixels representation and a back-projection scheme. Surface and background appearance models are learned online, leading to robust tracking in the presence of heavy occlusion and outliers. In both our tracking and reconstruction modules, the 3D object is implicitly embedded using a 3D level-set function. The framework is initialized with a simple shape primitive model (e.g. a sphere or a cube), and the real 3D object shape is tracked and reconstructed online. Unlike existing depth-based 3D reconstruction works, which either rely on calibrated/fixed camera set up or use the observed world map to track the depth camera, our framework can simultaneously track and reconstruct small moving objects. We use both qualitative and quantitative results to demonstrate the superior performance of both tracking and reconstruction of our method.

A novel framework for joint monocular 3D tracking and reconstruction is described that can handle untextured objects, occlusions, motion blur, changing background and imperfect lighting, and that can run at frame rate on a mobile phone. The method runs in parallel (i) level set based pose estimation and (ii) continuous max flow based shape optimisation. By avoiding a global computation of distance transforms typically used in level set methods, tracking rates here exceed 100Hz and 20Hz on a desktop and mobile phone, respectively, without needing a GPU. Tracking ambiguities are reduced by augmenting orientation information from the phone's inertial sensor. Reconstruction involves probabilistic integration of the 2D image statistics from keyframes into a 3D volume. Per-voxel posteriors are used instead of the standard likelihoods, giving increased accuracy and robustness. Shape coherency and compactness is then imposed using a total variational approach solved using globally optimal continuous max flow.

We propose a regression based learning framework that learns a set of shapes online, which can then be used to recover occluded object shapes. We represent shapes using their 2D discrete cosine transforms, and the key insight we propose is to regress low frequency harmonics, which represent the global properties of the shape, from high frequency harmonics, that encode the details of the object's shape. We learn the regression model using Locally Weighted Projection Regression (LWPR) which expedites online, incremental learning. After sufficient observation of a set of unoccluded shapes, the learned model can detect occlusion and recover the full shapes from the occluded ones. We demonstrate the ideas using a level-set based tracking system that provides shape and pose, however, the framework could be embedded in any segmentation-based tracking system. Our experiments demonstrate the efficacy of the method on a variety of objects using both real data and artificial data.

We propose a novel nonlinear, probabilistic and variational method for adding shape information to level set-based segmentation and tracking. Unlike previous work, we represent shapes with elliptic Fourier descriptors and learn their lower dimensional latent space using Gaussian Process Latent Variable Models. Segmentation is done by a nonlinear minimisation of an image-driven energy function in the learned latent space. We combine it with a 2D pose recovery stage, yielding a single, one shot, optimisation of both shape and pose. We demonstrate the performance of our method, both qualitatively and quantitatively, with multiple images, video sequences and latent spaces, capturing both shape kinematics and object class variance.

We propose a system for human computer interaction via 3D hand movements, based on a combination of visual tracking and a cheap, off-the-shelf, accelerometer. We use a 3D model and region based tracker, resulting in robustness to variations in illumination, motion blur and occlusions. At the same time the accelerometer allows us to deal with the multimodality in the silhouette to pose function. We synchronise the accelerometer and tracker online, by casting the calibration problem as a maximum covariance problem, which we then solve probabilistically. We show the effectiveness of our solution with multiple real-world tests and demonstration scenarios.

We propose a method for simultaneous shape-constrained segmentation and parameter recovery. The parameters can describe anything from 3D shape to 3D pose and we place no restriction on the topology of the shapes, i.e. they can have holes or be made of multiple parts. We use Shared Gaussian Process Latent Variable Models to learn multimodal shape-parameter spaces. These allow non-linear embeddings of the high-dimensional shape and parameter spaces in low dimensional spaces in a fully probabilistic manner. We propose a method for exploring the multimodality in the joint space in an efficient manner, by learning a mapping from the latent space to a space that encodes the similarity between shapes. We further extend the SGP-LVM to a model that makes use of a hierarchy of embeddings and show that this yields faster convergence and greater accuracy over the standard non-hierarchical embedding. Shapes are represented implicitly using level sets, and inference is made tractable by compressing the level set embedding functions with discrete cosine transforms. We show state of the art results in various fields, ranging from pose recovery to gaze tracking and to monocular 3D reconstruction.

Driver assistance helps save lives. Accurate 3D pose is required to establish if a traffic sign is relevant to the driver. We propose a real-time system that integrates single view detection with region-based 3D tracking of road signs. The optimal set of candidate detections is found, followed by AdaBoost cascades and SVMs. The 2D detections are then employed in simultaneous 2D segmentation and 3D pose tracking, using the known 3D model of the recognised traffic sign. We demonstrate the abilities of our system by tracking multiple road signs in real world scenarios.

We formulate a probabilistic framework for simultaneous 2D segmentation and 2D– 3D pose tracking, using a known 3D model (of arbitrary shape) of the segmented object. Our technique is region-based; at each frame we maximise the discrimination between statistical foreground and background models, by adjusting the pose parameters itera-tively. Unlike all previous work in 3D tracking, we use posterior membership proba-bilities for foreground and background pixels, rather than pixel likelihoods, and during periods of stable tracking we allow adaptation of the statistical foreground and back-ground models. We support our ideas with a real-time implementation, and use this to generate experimental results on both real and artificial video sequences, with a number of 3D models, to showcase the qualities of our tracker, and to demonstrate the benefit of using pixel-wise posteriors rather than likelihoods. 1