V-SysId identifies keypoints and 3D system from unlabeled videos.
problem Identifying keypoints and 3D system from unlabeled videos.
method Alternates between parameter estimation and extrinsic camera calibration, using motion equations as weak supervision.
result Utility of the approach demonstrated across various settings.
A method for camera calibration using heatmap regression for fisheye images.
problem Accurate and robust camera angle estimation from fisheye images in the Manhattan world.
method Heatmap regression to detect directions of labeled image coordinates, simultaneous rotation and fisheye distortion recovery.
result Our method outperforms conventional methods on large-scale datasets and with off-the-shelf cameras.
New approach learns traffic behavior from unlabelled video.
problem Learning from demonstration in settings with unlabelled raw video data.
method Calibrates camera, detects objects, tracks them, and uses trajectories for LfD.
result Can learn traffic behavior from raw video without additional expert knowledge.
The field of multiple view geometry has seen tremendous progress in reconstruction and calibration due to methods for extracting reliable point features and key developments in projective geometry. Point features, however, are not available in certain applications and result in unstructured point cloud reconstructions.…
Waymo Open Dataset provides a large, diverse, and synchronized LiDAR and camera dataset for autonomous driving research.
problem Limited diversity and scale in existing self-driving datasets hinder real-world problem alignment.
method Developed a new large-scale, high-quality, diverse dataset with synchronized LiDAR and camera data.
result The dataset is 15x more diverse than the largest existing dataset based on a proposed diversity metric.
Person re-identification (re-id), an emerging problem in visual surveillance, deals with maintaining entities of individuals whilst they traverse various locations surveilled by a camera network. From a visual perspective re-id is challenging due to significant changes in visual appearance of individuals in cameras wit…
Paper introduces deep learning for radar processing.
problem Lack of labeled radar data and need for expensive calibration.
method Uses deep learning on radar complex data, trains on calibration data, introduces radar augmentation.
result Superior performance on radar 4D detection task compared to classical methods.
New approach for camera-specific color constancy using few-shot meta-learning.
problem Domain gaps and lack of generalization across different cameras.
method Formulates color constancy as few-shot meta-learning tasks, leveraging annotated samples across different cameras.
result Significant reduction in data collection time and improved generalization to new cameras.
WoodScape dataset for fisheye cameras in autonomous driving.
problem Lack of public datasets for evaluating computer vision on fisheye images.
method Release WoodScape dataset with four surround view cameras and nine tasks.
result Encourages adaptation of computer vision models for fisheye cameras.
Study tackles open-set camera model identification, improving over state-of-the-art.
problem Identifying camera models from unknown ones in open-set scenarios.
method Feature extraction algorithms and classifiers for open-set recognition, evaluating different training protocols.
result A simple open-set training protocol yields the best results, improving over state-of-the-art solutions.
Deep learning system speeds up wildlife species identification from camera trap images.
problem Manual review of camera trap images is slow and resource-intensive.
method Combines machine and human intelligence for active learning.
result Matches state-of-the-art accuracy with minimal manual labels.
SoilingNet detects soiling on automotive cameras for better autonomous driving performance.
problem Soiling degrades the performance of automotive surround-view cameras, affecting autonomous driving.
method Created a new dataset and used a Convolutional Neural Network (CNN) architecture for soiling detection, combined with multi-task learning and data augmentation using GANs.
result Demonstrated high accuracy in soiling detection using CNN and multi-task learning.
Paper improves metrics for evaluating uncertainty in neural networks.
problem Improving accuracy of uncertainty estimation in neural networks.
method Proposed new metrics to address issues in existing quality metrics.
result Empirical results validate the superiority of the proposed metrics.
Deep learning identifies precipitation clouds from all-sky camera data.
problem Automating cloud warning systems for observatories.
method Deep learning using EfficientNet network.
result Average accuracy of 99% in identifying rainfall potential and 96% in cloud coverage.
New method corrects complex distortions in single view images.
problem Complex distortions in images, especially those caused by refractive surfaces.
method Differentiable image sampling and semantic information augmentation.
result Model can estimate and correct highly complex distortions.
Y-GAN uses multi-camera data to estimate depth maps without expensive hardware.
problem Depth perception for autonomous systems requires accurate 3D spatial information.
method Proposes Y-GAN, a deep convolutional generative adversarial network.
result Y-GAN estimates depth maps from multi-camera stereo images without ground truth data.
EvAn detects anomalies in event-based camera data with reduced complexity.
problem Anomaly detection in event-based camera data.
method Dual discriminator cGAN on event data with learned representation.
result Reduction in computational complexity compared to state-of-the-art methods.
Camera stickers can fool deep learning systems by manipulating the lens, achieving 49.6% misclassification rate.
problem The vulnerability of deep learning systems to physical adversarial attacks.
method Iterative procedure to update attack perturbation and threat model for physical realizability.
result Achieved 49.6% misclassification rate for targeted attacks on ImageNet classifiers.
A drone-based MOT algorithm tracks vehicles using neural network detections and TPMBM filter.
problem Tracking multiple vehicles from drone-mounted cameras.
method Neural network for object detection, TPMBM filter for trajectory estimation, von-Mises Fisher distribution for DOA.
result TPMBM filter optimally estimates vehicle trajectories.
Paper uses VAEs and GANs to estimate cryo-EM image orientation and camera parameters.
problem Estimating orientation and camera parameters from noisy cryo-EM images.
method Combines VAEs and GANs to learn latent representation, then designs estimation method.
result Geometric approach for fast cryo-EM biomolecule reconstruction.
Edge filters reduce video data transmission to datacenters.
problem Strain on wide area network infrastructure due to video camera deployments.
method FilterForward system with lightweight edge filters and microclassifiers.
result Reduces bandwidth use by an order of magnitude.
Self-supervised method estimates distances on fisheye cameras for autonomous driving.
problem Accurate Euclidean distance estimation on fisheye cameras for autonomous driving.
method Self-supervised scale-aware framework for monocular fisheye videos.
result State-of-the-art results on KITTI dataset, comparable to other methods.
A fast geometric regularizer improves event camera performance.
problem Event collapse in contrast maximization framework.
method Geometric regularizer to mitigate overfitting.
result State-of-the-art accuracy with reduced computational complexity.
New CNN approach reduces overconfidence in object classification predictions.
problem Overconfident predictions from deep models, especially SoftMax layer.
method Introduces CNN probabilistic approach using Logit layer for Bayesian inference.
result Proposed approach shows promising performance compared to SoftMax.
The introduction of cheap RGB-D cameras, stereo cameras, and LIDAR devices has given the computer vision community 3D information that conventional RGB cameras cannot provide. This data is often stored as a point cloud. In this paper, we present a novel method to apply the concept of convolutional neural networks to th…
Paper proposes a method to improve semantic segmentation for fisheye urban driving images.
problem Semantic segmentation for fisheye urban driving images is challenging due to distortion and lack of large datasets.
method A seven degrees of freedom augmentation method is proposed to transform rectilinear images into fisheye images.
result Training with seven-DoF augmentation improves model accuracy and robustness against distorted fisheye data.
In the emerging advancement in the branch of autonomous robotics, the ability of a robot to efficiently localize and construct maps of its surrounding is crucial. This paper deals with utilizing thermal-infrared cameras, as opposed to conventional cameras as the primary sensor to capture images of the robot's surroundi…
P3I learns holistic scene representations from a single image.
problem Inferring camera poses, object locations, and global scene structures from a single image.
method Combines search-based and gradient-based algorithms.
result P3I outperforms baselines on various image manipulation tasks.
Convolutional neural networks are commonly used to control the steering angle for autonomous cars. Most of the time, multiple long range cameras are used to generate lateral failure cases. In this paper we present a novel model to generate this data and label augmentation using only one short range fisheye camera. We p…
Collision avoidance is a critical task in many applications, such as ADAS (advanced driver-assistance systems), industrial automation and robotics. In an industrial automation setting, certain areas should be off limits to an automated vehicle for protection of people and high-valued assets. These areas can be quaranti…
Recent progress on many imaging and vision tasks has been driven by the use of deep feed-forward neural networks, which are trained by propagating gradients of a loss defined on the final output, back through the network up to the first layer that operates directly on the image. We propose back-propagating one step fur…
Our research focuses on analysing human activities according to a known behaviorist scenario, in case of noisy and high dimensional collected data. The data come from the monitoring of patients with dementia diseases by wearable cameras. We define a structural model of video recordings based on a Hidden Markov Model. N…
In recent years, many publications showed that convolutional neural network based features can have a superior performance to engineered features. However, not much effort was taken so far to extract local features efficiently for a whole image. In this paper, we present an approach to compute patch-based local feature…
Generative Map learns interpretable neural network maps for camera localization.
problem Creating interpretable maps for neural network-based camera localization.
method Combining generative models with Kalman filters and incorporating additional sensor information.
result Generative Map predicts images closely resembling the true scene and achieves comparable localization performance.
Can machine learning predict pedestrian destinations from video data?
problem Determining pedestrian destinations from video data without direct observation.
method Used density heatmaps as input to train a Random Forest predictor.
result Motivated further analysis of the methodology.
This work presents a new robust PCA method for foreground-background separation on freely moving camera video with possible dense and sparse corruptions. Our proposed method registers the frames of the corrupted video and then encodes the varying perspective arising from camera motion as missing data in a global model.…
We integrate camera pose correlations into deep models using Gaussian processes.
problem Lack of inter-frame reasoning in deep neural networks.
method Derive a principled framework combining camera pose information with deep models using a novel view kernel.
result Soft-prior knowledge aids pose-related vision tasks like novel view synthesis.
PIP-Net predicts pedestrian crossing intentions with up to 4-second lead.
problem Accurate pedestrian intention prediction for autonomous vehicles in real-world scenarios.
method Recurrent and temporal attention-based model using kinematic and spatial features.
result PIP-Net predicts pedestrian crossing intentions up to 4 seconds in advance.
This work presents a novel approach for robust PCA with total variation regularization for foreground-background separation and denoising on noisy, moving camera video. Our proposed algorithm registers the raw (possibly corrupted) frames of a video and then jointly processes the registered frames to produce a decomposi…
CNNs improve particle identification in ground-based gamma-ray astronomy.
problem Identifying particles in gamma-ray astronomy images.
method Used convolutional neural networks (CNNs) with PyTorch and TensorFlow.
result Improved accuracy in identifying gamma-rays and background particles.
Person re-identification (Re-ID) aims at matching images of the same person across disjoint camera views, which is a challenging problem in multimedia analysis, multimedia editing and content-based media retrieval communities. The major challenge lies in how to preserve similarity of the same person across video footag…
Automated video conferencing system improves user experience with ASD and VC.
problem Improve remote video conferencing experience through automated speaker detection and virtual cinematography.
method Uses 4K wide-FOV camera, depth camera, and microphone array to extract features and train machine learning models for ASD and VC.
result System performs within 0.3 MOS of an expert cinematographer, as rated by users.
Most existing machine learning classifiers are highly vulnerable to adversarial examples. An adversarial example is a sample of input data which has been modified very slightly in a way that is intended to cause a machine learning classifier to misclassify it. In many cases, these modifications can be so subtle that a …
End-to-end algorithm for joint object detection and sensor calibration from noisy images.
problem Jointly estimate objects and sensor parameters from noisy street-level imagery.
method Nested stochastic variational inference, soft data association, soft EM clustering, AD framework.
result Model is more robust to DNN misclassifications and generalizes across sign types.
Efficiently samples from large resolution images for faster inference.
problem Inferential and memory costs for large resolution images.
method Proposes a smaller proxy distribution to learn co-ordinates for regions of interest.
result Produces comparable results with ~10x faster inference and lower memory consumption.
New algorithm rSVDdpd improves robustness and scalability for video surveillance background modeling.
problem Camera tampering and noisy videos make background separation challenging.
method Introduces rSVDdpd, a robust singular value decomposition technique for scalable video surveillance.
result Demonstrates superior performance on benchmark and real-life datasets.
GATSBI uses GANs for SBI, improving posterior estimation in high dimensions.
problem Statistical inference on stochastic models without likelihoods.
method Adversarial approach to variational objective, amortized inference, implicit priors.
result GATSBI returns well-calibrated posterior estimates in high dimensions.
Novel framework assesses optical imaging hardware uncertainties.
problem Uncertainty in optical imaging modalities, especially ambiguity in parameter estimation.
method Invertible neural networks to map multispectral measurements to posterior probability distributions.
result Ambiguity in blood volume fraction estimation is a key finding.