This paper addresses the challenge of deploying salient object detection (SOD) on resource-constrained devices with real-time performance. While recent advances in deep neural networks have improved SOD, existing top-leading models are computationally expensive. We propose an efficient network design that combines traditional wisdom on SOD and the representation power of modern CNNs. Like biologically-inspired classical SOD methods relying on computing contrast cues to determine saliency of image regions, our model leverages Pixel Difference Convolutions (PDCs) to encode the feature contrasts. Differently, PDCs are incorporated in a CNN architecture so that the valuable contrast cues are extracted from rich feature maps. For efficiency, we introduce a difference convolution reparameterization (DCR) strategy that embeds PDCs into standard convolutions, eliminating computation and parameters at inference. Additionally, we introduce SpatioTemporal Difference Convolution (STDC) for video SOD, enhancing the standard 3D convolution with spatiotemporal contrast capture. Our models, SDNet for image SOD and STDNet for video SOD, achieve significant improvements in efficiency-accuracy trade-offs. On a Jetson Orin device, our models with $< $< 1M parameters operate at 46 FPS and 150 FPS on streamed images and videos, surpassing the second-best lightweight models in our experiments by more than $2\times$2× and $3\times$3× in speed with superior accuracy.

In the current era of generative AI breakthroughs, generating panoramic scenes from a single input image remains a key challenge. Most existing methods use diffusion-based iterative or simultaneous multi-view inpainting. However, the lack of global scene layout priors leads to subpar outputs with duplicated objects (e.g., multiple beds in a bedroom) or requires time-consuming human text inputs for each view. We propose L-MAGIC, a novel method leveraging large language models for guidance while diffusing multiple coherent views of 360° panoramic scenes. L-MAGIC harnesses pretrained diffusion and language models without fine-tuning, ensuring zero-shot performance. The output quality is further enhanced by super-resolution and multi-view fusion techniques. Extensive experiments demonstrate that the resulting panoramic scenes feature better scene layouts and perspective view rendering quality compared to related works, with >70% preference in human evaluations. Combined with conditional diffusion models, L-MAGIC can accept various input modalities, including but not limited to text, depth maps, sketches, and colored scripts. Applying depth estimation further enables 3D point cloud generation and dynamic scene exploration with fluid camera motion. Code is available at https://github.com/ZhipengCai/L-MAGIC-code-release.

LiDAR localization is a fundamental task in robotics and computer vision, which estimates the pose of a Li- DAR point cloud within a global map. Scene Coordinate Regression (SCR) has demonstrated state-of-the-art performance in this task. In SCR, a scene is represented as a neural network, which outputs the world coordinates for each point in the input point cloud. However, SCR treats all points equally during localization, ignoring the fact that not all objects are beneficial for localization. For exam-ple, dynamic objects and repeating structures often negatively impact SCR. To address this problem, we introduce LiSA, the first method that incorporates semantic aware-ness into SCR to boost the localization robustness and accuracy. To avoid extra computation or network parame-ters during inference, we distill the knowledge from a seg-mentation model to the original SCR network. Experi-ments show the superior performance of LiSA on standard LiDAR localization benchmarks compared to state-of-the- art methods. Applying knowledge distillation not only pre-serves high efficiency but also achieves higher localization accuracy than introducing extra semantic segmentation modules. We also analyze the benefit of semantic in-formation for LiDAR localization. Our code is released at https://github.com/Ybchun/LiSA.

Image matching is a fundamental computer vision problem. While learning-based methods achieve state-of-the-art performance on existing benchmarks, they generalize poorly to in-the-wild images. Such methods typically need to train separate models for different scene types (e.g., indoor vs. outdoor) and are impractical when the scene type is unknown in advance. One of the underlying problems is the limited scalability of existing data construction pipelines, which limits the diversity of standard image matching datasets. To address this problem, we propose GIM, a self-training framework for learning a single generalizable model based on any image matching architecture using internet videos, an abundant and diverse data source. Given an architecture, GIM first trains it on standard domain-specific datasets and then combines it with complementary matching methods to create dense labels on nearby frames of novel videos. These labels are filtered by robust fitting, and then enhanced by propagating them to distant frames. The final model is trained on propagated data with strong augmentations. Not relying on complex 3D reconstruction makes GIM much more efficient and less likely to fail than standard SfM-and-MVS based frameworks. We also propose ZEB, the first zero-shot evaluation benchmark for image matching. By mixing data from diverse domains, ZEB can thoroughly assess the cross-domain generalization performance of different methods. Experiments demonstrate the effectiveness and generality of GIM. Applying GIM consistently improves the zero-shot performance of 3 state-of-the-art image matching architectures as the number of downloaded videos increases (Fig. 1 (a)); with 50 hours of YouTube videos, the relative zero-shot performance improves by 6.9% − 18.1%. GIM also enables generalization to extreme cross-domain data such as Bird Eye View (BEV) images of projected 3D point clouds (Fig. 1 (c)). More importantly, our single zero-shot model consistently outperforms domain-specific baselines when evaluated on downstream tasks inherent to their respective domains. The code will be released upon acceptance.

This paper proposes a novel online evaluation protocol for Test Time Adaptation (TTA) methods, which penalizes slower methods by providing them with fewer samples for adaptation. TTA methods leverage unlabeled data at test time to adapt to distribution shifts. Though many effective methods have been proposed, their impressive performance usually comes at the cost of significantly increased computation budgets. Current evaluation protocols overlook the effect of this extra computation cost, affecting their real-world applicability. To address this issue, we propose a more realistic evaluation protocol for TTA methods, where data is received in an online fashion from a constant-speed data stream, thereby accounting for the method’s adaptation speed. We apply our proposed protocol to benchmark several TTA methods on multiple datasets and scenarios. Extensive experiments shows that, when accounting for inference speed, simple and fast approaches can outperform more sophisticated but slower methods. For example, SHOT from 2020, outperforms the state-of-the-art method SAR from 2023 under our online setting. Our results reveal the importance of developing practical TTA methods that are both accurate and efficient.

We introduce OpenBot-Fleet, a comprehensive open-source cloud robotics system for navigation. OpenBot-Fleet uses smartphones for sensing, local compute and communication, Google Firebase for secure cloud storage and off-board compute, and a robust yet low-cost wheeled robot to act in real-world environments. The robots collect task data and upload it to the cloud where navigation policies can be learned either offline or online and can then be sent back to the robot fleet. In our experiments we distribute 72 robots to a crowd of workers who operate them in homes, and show that OpenBot-Fleet can learn robust navigation policies that generalize to unseen homes with >80% success rate. OpenBot-Fleet represents a significant step forward in cloud robotics, making it possible to deploy large continually learning robot fleets in a cost-effective and scalable manner. All materials can be found at https://www.openbot.org/.

Online continual learning, the process of training models on streaming data, has gained increasing attention in recent years. However, a critical aspect often overlooked is the label delay, where new data may not be labeled due to slow and costly annotation processes. We introduce a new continual learning framework with explicit modeling of the label delay between data and label streams over time steps. In each step, the framework reveals both unlabeled data from the current time step t and labels delayed with d steps, from the time step t−d. In our extensive experiments amounting to 1060 GPU days, we show that merely augmenting the computational resources is insufficient to tackle this challenge. Our findings underline a notable performance decline when solely relying on labeled data when the label delay becomes significant. More surprisingly, when using state-of-the-art SSL and TTA techniques to utilize the newer, unlabeled data, they fail to surpass the performance of a naïve method that simply trains on the delayed supervised stream. To this end, we introduce a simple, efficient baseline that rehearses from the labeled memory samples that are most similar to the new unlabeled samples. This method bridges the accuracy gap caused by label delay without significantly increasing computational complexity. We show experimentally that our method is the least affected by the label delay factor and in some cases successfully recovers the accuracy of the non-delayed counterpart. We conduct various ablations and sensitivity experiments, demonstrating the effectiveness of our approach.

Humans naturally exploit haptic feedback during contact-rich tasks like loading a dishwasher or stocking a bookshelf. Current robotic systems focus on avoiding unexpected contact, often relying on strategically placed environment sensors. Recently, contact-exploiting manipulation policies have been trained in simulation and deployed on real robots. However, they require some form of real-world adaptation to bridge the sim-to-real gap, which might not be feasible in all scenarios. In this paper we train a contact-exploiting manipulation policy in simulation for the contact-rich household task of loading plates into a slotted holder, which transfers without any fine-tuning to the real robot. We investigate various factors necessary for this zero-shot transfer, like time delay modeling, memory representation, and domain randomization. Our policy transfers with minimal sim-to-real gap and significantly outperforms heuristic and learnt baselines. It also generalizes well to a cup and plates of different sizes and weights. The project website is https://sites.google.com/view/compliant-object-insertion.

Control design for robotic systems is complex and often requires solving an optimization to follow a trajectory accurately. Online optimization approaches like Model Predictive Control (MPC) have been shown to achieve great tracking performance, but require high computing power. Conversely, learning-based offline optimization approaches, such as Reinforcement Learning (RL), allow fast and efficient execution on the robot but hardly match the accuracy of MPC in trajectory tracking tasks. In systems with limited compute, such as aerial vehicles, an accurate controller that is efficient at execution time is imperative. We propose an Analytic Policy Gradient (APG) method to tackle this problem. APG exploits the availability of differentiable simulators by training a controller offline with gradient descent on the tracking error. We address training instabilities that frequently occur with APG through curriculum learning and experiment on a widely used controls benchmark, the CartPole, and two common aerial robots, a quadrotor and a fixed-wing drone. Our proposed method outperforms both model-based and model-free RL methods in terms of tracking error. Concurrently, it achieves similar performance to MPC while requiring more than an order of magnitude less computation time. Our work provides insights into the potential of APG as a promising control method for robotics. To facilitate the exploration of APG, we open-source our code and make it available atgithub.com/lis-epfl/apg_trajectory_tracking.

We present a visual-inertial depth estimation pipeline that integrates monocular depth estimation and visual- inertial odometry to produce dense depth estimates with metric scale. Our approach performs global scale and shift alignment against sparse metric depth, followed by learning-based dense alignment. We evaluate on the TartanAir and VOID datasets, observing up to 30% reduction in inverse RMSE with dense scale alignment relative to performing just global alignment alone. Our approach is especially competitive at low density; with just 150 sparse metric depth points, our dense- to-dense depth alignment method achieves over 50 % lower iRMSE over sparse-to-dense depth completion by KBNet, currently the state of the art on VOID. We demonstrate successful zero-shot transfer from synthetic TartanAir to real-world VOID data and perform generalization tests on NYUv2 and VCU-RVI. Our approach is modular and is compatible with a variety of monocular depth estimation models.

Event cameras have emerged as a promising vision sensor in recent years due to their unparalleled temporal resolution and dynamic range. While registration of 2D RGB images to 3D point clouds is a long-standing problem in computer vision, no prior work studies 2D-3D registration for event cameras. To this end, we propose E2PNet, the first learning-based method for event-to-point cloud registration. The core of E2PNet is a novel feature representation network called Event-Points-to-Tensor (EP2T), which encodes event data into a 2D grid-shaped feature tensor. This grid-shaped feature enables matured RGB-based frameworks to be easily used for event-to-point cloud registration, without changing hyper-parameters and the training procedure. EP2T treats the event input as spatio-temporal point clouds. Unlike standard 3D learning architectures that treat all dimensions of point clouds equally, the novel sampling and information aggregation modules in EP2T are designed to handle the inhomogeneity of the spatial and temporal dimensions. Experiments on the MVSEC and VECtor datasets demonstrate the superiority of E2PNet over hand-crafted and other learning-based methods. Compared to RGB-based registration, E2PNet is more robust to extreme illumination or fast motion due to the use of event data. Beyond 2D-3D registration, we also show the potential of EP2T for other vision tasks such as flow estimation, event-to-image reconstruction and object recognition. The source code can be found at: https://github.com/Xmu-qcj/E2PNet.

Convolutional neural networks (CNNs) have been very successful at solving a variety of computer vision tasks such as object classification and detection, semantic segmentation, activity understanding, to name just a few. One key enabling factor for their great performance has been the ability to train very deep networks. Despite their huge success in many tasks, CNNs do not work well with non-euclidean data, which is prevalent in many real-world applications. Graph Convolutional Networks (GCNs) offer an alternative that allows for non-Eucledian data input to a neural network. While GCNs already achieve encouraging results, they are currently limited to architectures with a relatively small number of layers, primarily due to vanishing gradients during training. This work transfers concepts such as residual/dense connections and dilated convolutions from CNNs to GCNs in order to successfully train very deep GCNs. We show the benefit of using deep GCNs (with as many as 112 layers) experimentally across various datasets and tasks. Specifically, we achieve very promising performance in part segmentation and semantic segmentation on point clouds and in node classification of protein functions across biological protein-protein interaction (PPI) graphs. We believe that the insights in this work will open avenues for future research on GCNs and their application to further tasks not explored in this paper. The source code for this work is available at https://github.com/lightaime/deep_gcns_torch and https://github.com/lightaime/deep_gcns for PyTorch and TensorFlow implementations respectively.

Fixed-wing drones must navigate to the desired location accurately for maneuvers such as picking up objects and perching. However, current GNSS receivers limit their navigation accuracy to several meters in outdoor environments, making such maneuvers impossible. RTK GNSS can improve flight accuracy, but it requires ground stations at the target location and additional communication modules on the drone. Here, we describe a fixed-wing platform with onboard computation that uses positional information from a GNSS receiver and vision from an onboard camera. The drone relies on a GNSS signal for flying towards a point of interest and switches to vision-based information to accurately reach the target. We conducted outdoor experiments to compare the flight accuracy of three navigation methods: GNSS, RTK GNSS, and the proposed GNSS-vision method. We also systematically assessed the robustness of vision-based control to compensate for GNSS errors and quantify the accuracy of the proposed method. Our results show that the accuracy of the proposed GNSS-vision system is on par with RTK GNSS. GNSS-vision reduces the average error of GNSS by over an order of magnitude, from 3.033 m to 0.283 m, and reduces the variance across repeated flights from 2.095 m to 0.309 m. We open-source the software-hardware architecture used in this paper to enable the research community to build on these results and expand the capabilities of fixed-wing drones.

Deep graph neural networks (GNNs) have achieved excellent results on various tasks on increasingly large graph datasets with millions of nodes and edges. However, memory complexity has become a major obstacle when training deep GNNs for practical applications due to the immense number of nodes, edges, and intermediate activations. To improve the scalability of GNNs, prior works propose smart graph sampling or partitioning strategies to train GNNs with a smaller set of nodes or sub-graphs. In this work, we study reversible connections, group convolutions, weight tying, and equilibrium models to advance the memory and parameter efficiency of GNNs. We find that reversible connections in combination with deep network architectures enable the training of overparameterized GNNs that significantly outperform existing methods on multiple datasets. Our models RevGNN-Deep (1001 layers with 80 channels each) and RevGNN-Wide (448 layers with 224 channels each) were both trained on a single commodity GPU and achieve an ROC-AUC of 87.74 $\pm$ 0.13 and 88.14 $\pm$ 0.15 on the ogbn-proteins dataset. To the best of our knowledge, RevGNN-Deep is the deepest GNN in the literature by one order of magnitude.

Current robots are either expensive or make significant compromises on sensory richness, computational power, and communication capabilities. We propose to leverage smartphones to equip robots with extensive sensor suites, powerful computational abilities, state-of-the-art communication channels, and access to a thriving software ecosystem. We design a small electric vehicle that costs $50 and serves as a robot body for standard Android smartphones. We develop a software stack that allows smartphones to use this body for mobile operation and demonstrate that the system is sufficiently powerful to support advanced robotics workloads such as person following and real-time autonomous navigation in unstructured environments. Controlled experiments demonstrate that the presented approach is robust across different smartphones and robot bodies.

Acrobatic flight with quadrotors is extremely challenging. Maneuvers such as the loop, matty flip, or barrel roll require high thrust and extreme angular accelerations that push the platform to its limits. Human drone pilots require years of practice to safely master such maneuvers. Yet, a tiny mistake could make the platform lose control, and brutally crash. This short paper describes an approach to safely train acrobatic controllers in simulation and deploy them with no fine-tuning zero-shot transfer on physical quadrotors. The approach uses only onboard sensing and computation.

One major factor impeding more widespread adoption of deep neural networks (DNNs) is their lack of robustness, which is essential for safety-critical applications such as autonomous driving. This has motivated much recent work on adversarial attacks for DNNs, which mostly focus on pixel-level perturbations void of semantic meaning. In contrast, we present a general framework for adversarial attacks on trained agents, which covers semantic perturbations to the environment of the agent performing the task as well as pixel-level attacks. To do this, we re-frame the adversarial attack problem as learning a distribution of parameters that always fools the agent. In the semantic case, our proposed adversary (denoted as BBGAN) is trained to sample parameters that describe the environment with which the black-box agent interacts, such that the agent performs its dedicated task poorly in this environment. We apply BBGAN on three different tasks, primarily targeting aspects of autonomous navigation: object detection, self-driving, and autonomous UAV racing. On these tasks, BBGAN can generate failure cases that consistently fool a trained agent.

Architecture design has become a crucial component of successful deep learning. Recent progress in automatic neural architecture search (NAS) shows a lot of promise. However, discovered architectures often fail to generalize in the final evaluation. Architectures with a higher validation accuracy during the search phase may perform worse in the evaluation. Aiming to alleviate this common issue, we introduce sequential greedy architecture search (SGAS), an efficient method for neural architecture search. By dividing the search procedure into sub-problems, SGAS chooses and prunes candidate operations in a greedy fashion. We apply SGAS to search architectures for Convolutional Neural Networks (CNN) and Graph Convolutional Networks (GCN). Extensive experiments show that SGAS is able to find state-of-the-art architectures for tasks such as image classification, point cloud classification and node classification in protein-protein interaction graphs with minimal computational cost.

Convolutional Neural Networks (CNNs) achieve impressive performance in a wide variety of fields. Their success benefited from a massive boost when very deep CNN models were able to be reliably trained. Despite their merits, CNNs fail to properly address problems with non-Euclidean data. To overcome this challenge, Graph Convolutional Networks (GCNs) build graphs to represent non-Euclidean data, borrow concepts from CNNs, and apply them in training. GCNs show promising results, but they are usually limited to very shallow models due to the vanishing gradient problem. As a result, most state-of-the-art GCN models are no deeper than 3 or 4 layers. In this work, we present new ways to successfully train very deep GCNs. We do this by borrowing concepts from CNNs, specifically residual/dense connections and dilated convolutions, and adapting them to GCN architectures. Extensive experiments show the positive effect of these deep GCN frameworks. Finally, we use these new concepts to build a very deep 56-layer GCN, and show how it significantly boosts performance (+3.7% mIoU over state-of-the-art) in the task of point cloud semantic segmentation. We believe that the community can greatly benefit from this work, as it opens up many opportunities for advancing GCN-based research.

Deep networks trained on demonstrations of human driving have learned to follow roads and avoid obstacles. However, driving policies trained via imitation learning cannot be controlled at test time. A vehicle trained end-to-end to imitate an expert cannot be guided to take a specific turn at an upcoming intersection. This limits the utility of such systems. We propose to condition imitation learning on high-level command input. At test time, the learned driving policy functions as a chauffeur that handles sensorimotor coordination but continues to respond to navigational commands. We evaluate different architectures for conditional imitation learning in vision-based driving. We conduct experiments in realistic three-dimensional simulations of urban driving and on a 1/5 scale robotic truck that is trained to drive in a residential area. Both systems drive based on visual input yet remain responsive to high-level navigational commands.

Correlation filter (CF) based trackers have recently gained a lot of popularity due to their impressive performance on benchmark datasets, while maintaining high frame rates. A significant amount of recent research focuses on the incorporation of stronger features for a richer representation of the tracking target. However, this only helps to discriminate the target from background within a small neighborhood. In this paper, we present a framework that allows the explicit incorporation of global context within CF trackers. We reformulate the original optimization problem and provide a closed form solution for single and multi-dimensional features in the primal and dual domain. Extensive experiments demonstrate that this framework significantly improves the performance of many CF trackers with only a modest impact on frame rate.

In this paper, we propose a persistent, robust and autonomous object tracking system for unmanned aerial vehicles (UAVs) called Persistent Aerial Tracking (PAT). A computer vision and control strategy is applied to a diverse set of moving objects (e.g. humans, animals, cars, boats, etc.) integrating multiple UAVs with a stabilized RGB camera. A novel strategy is employed to successfully track objects over a long period, by `handing over the camera' from one UAV to another. We evaluate several state-of-the-art trackers on the VIVID aerial video dataset and additional sequences that are specifically tailored to low altitude UAV target tracking. Based on the evaluation, we select the leading tracker and improve upon it by optimizing for both speed and performance, integrate the complete system into an off-the-shelf UAV, and obtain promising results showing the robustness of our solution in real-world aerial scenarios.