Search

We present results from a combined radio polarization and emission line study of five type 2 quasars at $z<0.2$ with the Karl G. Jansky Very Large Array (VLA) B-array at 5 GHz and Hubble Space Telescope (HST) [O III] observations. These five sources are known to exhibit close association between radio structures and ionized gas morphology and kinematics. Four sources (J0945+1737, J1000+1242, J1356+1026 and J1430+1339) show polarization in the current data. J1010+1413 is the unpolarized source in our sample. We detect $0.5-1\%$ fractional polarization in the radio cores and a high fractional polarization ($10-30\%$) in the lobes of these sources. The morphological, spectral and polarization properties suggest a jet origin for radio emission in J0945+1737, J1000+1242, J1010+1413 and J1430+1339 whereas the current data cannot fully discern the origin of radio emission (jet or wind) in J1356+1026. An anti-correlation between various polarized knots in the radio and [O III] emission is observed in our sources, similar to that observed in some radio-loud AGN in the literature. This suggests that the radio emission is likely to be depolarized by the emission-line gas. By modeling the depolarization effects, we estimate the size of the emission-line gas clouds to be $\sim(2.8\pm1.7)\times10^{-5}$ parsec and the amount of thermal material mixed with the synchrotron plasma to be $\sim(1.01\pm0.08)\times10^{6}$M$_\odot$ in the lobe of J0945+1737 (which exhibits the most prominent polarization signature in its lobe). The current work demonstrates that the interplay of jets/winds and emission-line gas is most likely responsible for the nature of radio outflows in radio-quiet AGN., Comment: 17 pages, 8 figures; v2: matches published version

Prior work has studied different visual modalities in isolation and developed separate architectures for recognition of images, videos, and 3D data. Instead, in this paper, we propose a single model which excels at classifying images, videos, and single-view 3D data using exactly the same model parameters. Our 'Omnivore' model leverages the flexibility of transformer-based architectures and is trained jointly on classification tasks from different modalities. Omnivore is simple to train, uses off-the-shelf standard datasets, and performs at-par or better than modality-specific models of the same size. A single Omnivore model obtains 86.0% on ImageNet, 84.1% on Kinetics, and 67.1% on SUN RGB-D. After finetuning, our models outperform prior work on a variety of vision tasks and generalize across modalities. Omnivore's shared visual representation naturally enables cross-modal recognition without access to correspondences between modalities. We hope our results motivate researchers to model visual modalities together., Comment: Accepted at CVPR 2022 (Oral Presentation)

Current object detectors are limited in vocabulary size due to the small scale of detection datasets. Image classifiers, on the other hand, reason about much larger vocabularies, as their datasets are larger and easier to collect. We propose Detic, which simply trains the classifiers of a detector on image classification data and thus expands the vocabulary of detectors to tens of thousands of concepts. Unlike prior work, Detic does not need complex assignment schemes to assign image labels to boxes based on model predictions, making it much easier to implement and compatible with a range of detection architectures and backbones. Our results show that Detic yields excellent detectors even for classes without box annotations. It outperforms prior work on both open-vocabulary and long-tail detection benchmarks. Detic provides a gain of 2.4 mAP for all classes and 8.3 mAP for novel classes on the open-vocabulary LVIS benchmark. On the standard LVIS benchmark, Detic obtains 41.7 mAP when evaluated on all classes, or only rare classes, hence closing the gap in performance for object categories with few samples. For the first time, we train a detector with all the twenty-one-thousand classes of the ImageNet dataset and show that it generalizes to new datasets without finetuning. Code is available at \url{https://github.com/facebookresearch/Detic}., Comment: ECCV 2022 camera ready. Code is available at https://github.com/facebookresearch/Detic

We present a study of a luminous, z=0.15, type-2 quasar (log [L([OIII])/(erg/s)]=42.8) from the Quasar Feedback Survey. It is classified as 'radio-quiet' (log [L(1.4 GHz)/(W/Hz)]=23.8); however, radio imaging reveals ~1 kpc low-power jets (log [Pjet/(erg/s)]=44) inclined into the plane of the galaxy disk. We combine MUSE and ALMA observations to map stellar kinematics and ionised and molecular gas properties. The jets are seen to drive galaxy-wide bi-conical turbulent outflows, reaching W80 = 1000-1300 km/s, in the ionised phase (traced via optical emission-lines), which also have increased electron densities compared to the quiescent gas. The turbulent gas is driven perpendicular to the jet axis and is escaping along the galaxy minor axis, reaching 7.5 kpc on both sides. Traced via CO(3-2) emission, the turbulent material in molecular gas phase is one-third as spatially extended and has 3 times lower velocity-dispersion as compared to ionised gas. The jets are seen to be strongly interacting with the interstellar medium (ISM) through enhanced ionised emission and disturbed/depleted molecular gas at the jet termini. We see further evidence for jet-induced feedback through significantly higher stellar velocity-dispersion aligned, and co-spatial with, the jet axis (<5 deg). We discuss possible negative and positive feedback scenarios arising due to the interaction of the low-power jets with the ISM in the context of recent jet-ISM interaction simulations, which qualitatively agree with our observations. We discuss how jet-induced feedback could be an important feedback mechanism even in bolometrically luminous 'radio-quiet' quasars., Comment: Accepted for publication in MNRAS. Main manuscript has 21 pages with 8 figures. Supplementary material is available for download under "Ancillary files" or by downloading the source file listed under "Other formats"

We find Mask2Former also achieves state-of-the-art performance on video instance segmentation without modifying the architecture, the loss or even the training pipeline. In this report, we show universal image segmentation architectures trivially generalize to video segmentation by directly predicting 3D segmentation volumes. Specifically, Mask2Former sets a new state-of-the-art of 60.4 AP on YouTubeVIS-2019 and 52.6 AP on YouTubeVIS-2021. We believe Mask2Former is also capable of handling video semantic and panoptic segmentation, given its versatility in image segmentation. We hope this will make state-of-the-art video segmentation research more accessible and bring more attention to designing universal image and video segmentation architectures., Comment: Code and models: https://github.com/facebookresearch/Mask2Former

We study the ground-state properties of ferromagnetic quasi-one-dimensional quantum wires using the quantum Monte Carlo (QMC) method for various wire widths $b$ and density parameters $r_\text{s}$. The correlation energy, pair-correlation function, static structure factor, and momentum density are calculated at high density, $r_\text{s}=0.5$. It is observed that the peak in the static structure factor at $k=2k_\text{F}$ grows as the wire width decreases. We obtain the Tomonaga-Luttinger liquid parameter $K_\rho$ from the momentum density. It is found that $K_\rho$ increases by about $10$\% between wire widths $b=0.01$ and $b=0.5$. We also obtain ground-state properties of finite thickness wires theoretically using the first-order random phase approximation (RPA) with exchange and self-energy contributions, which is exact in the high-density limit. Analytical expressions for the static structure factor and correlation energy are derived for $b \ll r_\text{s}<1$. It is found that the correlation energy varies as $b^2$ for $b \ll r_\text{s}$ from its value for an infinitely thin wire. It is observed that the correlation energy depends significantly on the wire model used (harmonic versus cylindrical confinement). The first-order RPA expressions for the structure factor, pair-correlation function, and correlation energy are numerically evaluated for several values of $b$ and $r_\text{s} \leq 1$. These are compared with the QMC results in the range of applicability of the theory.

Image segmentation is about grouping pixels with different semantics, e.g., category or instance membership, where each choice of semantics defines a task. While only the semantics of each task differ, current research focuses on designing specialized architectures for each task. We present Masked-attention Mask Transformer (Mask2Former), a new architecture capable of addressing any image segmentation task (panoptic, instance or semantic). Its key components include masked attention, which extracts localized features by constraining cross-attention within predicted mask regions. In addition to reducing the research effort by at least three times, it outperforms the best specialized architectures by a significant margin on four popular datasets. Most notably, Mask2Former sets a new state-of-the-art for panoptic segmentation (57.8 PQ on COCO), instance segmentation (50.1 AP on COCO) and semantic segmentation (57.7 mIoU on ADE20K)., Comment: CVPR 2022. Project page/code/models: https://bowenc0221.github.io/mask2former

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of daily-life activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/, Comment: To appear in the Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022. This version updates the baseline result numbers for the Hands and Objects benchmark (appendix)

We propose 3DETR, an end-to-end Transformer based object detection model for 3D point clouds. Compared to existing detection methods that employ a number of 3D-specific inductive biases, 3DETR requires minimal modifications to the vanilla Transformer block. Specifically, we find that a standard Transformer with non-parametric queries and Fourier positional embeddings is competitive with specialized architectures that employ libraries of 3D-specific operators with hand-tuned hyperparameters. Nevertheless, 3DETR is conceptually simple and easy to implement, enabling further improvements by incorporating 3D domain knowledge. Through extensive experiments, we show 3DETR outperforms the well-established and highly optimized VoteNet baselines on the challenging ScanNetV2 dataset by 9.5%. Furthermore, we show 3DETR is applicable to 3D tasks beyond detection, and can serve as a building block for future research., Comment: Accepted at ICCV 2021

We propose Anticipative Video Transformer (AVT), an end-to-end attention-based video modeling architecture that attends to the previously observed video in order to anticipate future actions. We train the model jointly to predict the next action in a video sequence, while also learning frame feature encoders that are predictive of successive future frames' features. Compared to existing temporal aggregation strategies, AVT has the advantage of both maintaining the sequential progression of observed actions while still capturing long-range dependencies--both critical for the anticipation task. Through extensive experiments, we show that AVT obtains the best reported performance on four popular action anticipation benchmarks: EpicKitchens-55, EpicKitchens-100, EGTEA Gaze+, and 50-Salads; and it wins first place in the EpicKitchens-100 CVPR'21 challenge., Comment: ICCV 2021. Ranked #1 in CVPR'21 EPIC-Kitchens-100 Action Anticipation challenge. Webpage/code/models: http://facebookresearch.github.io/AVT

We introduce WyPR, a Weakly-supervised framework for Point cloud Recognition, requiring only scene-level class tags as supervision. WyPR jointly addresses three core 3D recognition tasks: point-level semantic segmentation, 3D proposal generation, and 3D object detection, coupling their predictions through self and cross-task consistency losses. We show that in conjunction with standard multiple-instance learning objectives, WyPR can detect and segment objects in point cloud data without access to any spatial labels at training time. We demonstrate its efficacy using the ScanNet and S3DIS datasets, outperforming prior state of the art on weakly-supervised segmentation by more than 6% mIoU. In addition, we set up the first benchmark for weakly-supervised 3D object detection on both datasets, where WyPR outperforms standard approaches and establishes strong baselines for future work., Comment: CVPR 2021

We present CREST, an approach for causal reasoning in simulation to learn the relevant state space for a robot manipulation policy. Our approach conducts interventions using internal models, which are simulations with approximate dynamics and simplified assumptions. These interventions elicit the structure between the state and action spaces, enabling construction of neural network policies with only relevant states as input. These policies are pretrained using the internal model with domain randomization over the relevant states. The policy network weights are then transferred to the target domain (e.g., the real world) for fine tuning. We perform extensive policy transfer experiments in simulation for two representative manipulation tasks: block stacking and crate opening. Our policies are shown to be more robust to domain shifts, more sample efficient to learn, and scale to more complex settings with larger state spaces. We also show improved zero-shot sim-to-real transfer of our policies for the block stacking task., Comment: Presented at the 2021 IEEE International Conference on Robotics and Automation (ICRA 2021). This version contains the camera-ready paper with supplementary materials and minor typo fixes. Project page: https://sites.google.com/view/crest-causal-struct-xfer-manip

We present a solution to multi-robot distributed semantic mapping of novel and unfamiliar environments. Most state-of-the-art semantic mapping systems are based on supervised learning algorithms that cannot classify novel observations online. While unsupervised learning algorithms can invent labels for novel observations, approaches to detect when multiple robots have independently developed their own labels for the same new class are prone to erroneous or inconsistent matches. These issues worsen as the number of robots in the system increases and prevent fusing the local maps produced by each robot into a consistent global map, which is crucial for cooperative planning and joint mission summarization. Our proposed solution overcomes these obstacles by having each robot learn an unsupervised semantic scene model online and use a multiway matching algorithm to identify consistent sets of matches between learned semantic labels belonging to different robots. Compared to the state of the art, the proposed solution produces 20-60% higher quality global maps that do not degrade even as many more local maps are fused., Comment: 7 pages, 6 figures, 1 table; accepted for presentation in IEEE Int. Conf. on Robotics and Automation, ICRA '21, Xi'an, China, June 2021

We present the first results from the Quasar Feedback Survey, a sample of 42 z<0.2, [O III] luminous AGN (L[O III]>10^42.1 ergs/s) with moderate radio luminosities (i.e. L(1.4GHz)>10^23.4 W/Hz; median L(1.4GHz)=5.9x10^23 W/Hz). Using high spatial resolution (~0.3-1 arcsec), 1.5-6 GHz radio images from the Very Large Array, we find that 67 percent of the sample have spatially extended radio features, on ~1-60 kpc scales. The radio sizes and morphologies suggest that these may be lower radio luminosity versions of compact, radio-loud AGN. By combining the radio-to-infrared excess parameter, spectral index, radio morphology and brightness temperature, we find radio emission in at least 57 percent of the sample that is associated with AGN-related processes (e.g. jets, quasar-driven winds or coronal emission). This is despite only 9.5-21 percent being classified as radio-loud using traditional criteria. The origin of the radio emission in the remainder of the sample is unclear. We find that both the established anti-correlation between radio size and the width of the [O III] line, and the known trend for the most [O III] luminous AGN to be associated with spatially-extended radio emission, also hold for our sample of moderate radio luminosity quasars. These observations add to the growing evidence of a connection between the radio emission and ionised gas in quasar host galaxies. This work lays the foundation for deeper investigations into the drivers and impact of feedback in this unique sample., Comment: Accepted for publication in MNRAS. Data products from this paper and the survey pilot papers are available through our website: https://blogs.ncl.ac.uk/quasarfeedbacksurvey/. The extensive supplementary material (containing additional figures and information on individual targets) is available for download under "Ancillary files" or by downloading the source file listed under "Other formats"

A common approach to solving physical reasoning tasks is to train a value learner on example tasks. A limitation of such an approach is that it requires learning about object dynamics solely from reward values assigned to the final state of a rollout of the environment. This study aims to address this limitation by augmenting the reward value with self-supervised signals about object dynamics. Specifically, we train the model to characterize the similarity of two environment rollouts, jointly with predicting the outcome of the reasoning task. This similarity can be defined as a distance measure between the trajectory of objects in the two rollouts, or learned directly from pixels using a contrastive formulation. Empirically, we find that this approach leads to substantial performance improvements on the PHYRE benchmark for physical reasoning (Bakhtin et al., 2019), establishing a new state-of-the-art., Comment: ICML 2021 Workshop on Self-Supervised Learning for Reasoning and Perception; Webpage/Code: https://facebookresearch.github.io/DynamicsAware

Pretraining on large labeled datasets is a prerequisite to achieve good performance in many computer vision tasks like 2D object recognition, video classification etc. However, pretraining is not widely used for 3D recognition tasks where state-of-the-art methods train models from scratch. A primary reason is the lack of large annotated datasets because 3D data is both difficult to acquire and time consuming to label. We present a simple self-supervised pertaining method that can work with any 3D data - single or multiview, indoor or outdoor, acquired by varied sensors, without 3D registration. We pretrain standard point cloud and voxel based model architectures, and show that joint pretraining further improves performance. We evaluate our models on 9 benchmarks for object detection, semantic segmentation, and object classification, where they achieve state-of-the-art results and can outperform supervised pretraining. We set a new state-of-the-art for object detection on ScanNet (69.0% mAP) and SUNRGBD (63.5% mAP). Our pretrained models are label efficient and improve performance for classes with few examples.