Autonomous flying robots, e.g. multirotors, often rely on a neural network that makes predictions based on a camera image. These deep learning (DL) models can compute surprising results if applied to input images outside the training domain. Adversarial attacks exploit this fault, for example, by computing small images, so-called adversarial patches, that can be placed in the environment to manipulate the neural network's prediction. We introduce flying adversarial patches, where an image is mounted on another flying robot and therefore can be placed anywhere in the field of view of a victim multirotor. For an effective attack, we compare three methods that simultaneously optimize the adversarial patch and its position in the input image. We perform an empirical validation on a publicly available DL model and dataset for autonomous multirotors. Ultimately, our attacking multirotor would be able to gain full control over the motions of the victim multirotor.
Heatmap-based anatomical landmark detection is still facing two unresolved challenges: 1) inability to accurately evaluate the distribution of heatmap; 2) inability to effectively exploit global spatial structure information. To address the computational inability challenge, we propose a novel position-aware and sample-aware central loss. Specifically, our central loss can absorb position information, enabling accurate evaluation of the heatmap distribution. More advanced is that our central loss is sample-aware, which can adaptively distinguish easy and hard samples and make the model more focused on hard samples while solving the challenge of extreme imbalance between landmarks and non-landmarks. To address the challenge of ignoring structure information, a Coordinated Transformer, called CoorTransformer, is proposed, which establishes long-range dependencies under the guidance of landmark coordination information, making the attention more focused on the sparse landmarks while taking advantage of global spatial structure. Furthermore, CoorTransformer can speed up convergence, effectively avoiding the defect that Transformers have difficulty converging in sparse representation learning. Using the advanced CoorTransformer and central loss, we propose a generalized detection model that can handle various scenarios, inherently exploiting the underlying relationship between landmarks and incorporating rich structural knowledge around the target landmarks. We analyzed and evaluated CoorTransformer and central loss on three challenging landmark detection tasks. The experimental results show that our CoorTransformer outperforms state-of-the-art methods, and the central loss significantly improves the performance of the model with p-values< 0.05.
Existing vector quantization (VQ) based autoregressive models follow a two-stage generation paradigm that first learns a codebook to encode images as discrete codes, and then completes generation based on the learned codebook. However, they encode fixed-size image regions into fixed-length codes and ignore their naturally different information densities, which results in insufficiency in important regions and redundancy in unimportant ones, and finally degrades the generation quality and speed. Moreover, the fixed-length coding leads to an unnatural raster-scan autoregressive generation. To address the problem, we propose a novel two-stage framework: (1) Dynamic-Quantization VAE (DQ-VAE) which encodes image regions into variable-length codes based on their information densities for an accurate and compact code representation. (2) DQ-Transformer which thereby generates images autoregressively from coarse-grained (smooth regions with fewer codes) to fine-grained (details regions with more codes) by modeling the position and content of codes in each granularity alternately, through a novel stacked-transformer architecture and shared-content, non-shared position input layers designs. Comprehensive experiments on various generation tasks validate our superiorities in both effectiveness and efficiency. Code will be released at https://github.com/CrossmodalGroup/DynamicVectorQuantization.
The web is a treasure trove for data that is increasingly used by computer scientists for building large machine learning models as well as non-computer scientists for social studies or marketing analyses. As such, web-crawling is an essential tool for both computational and non-computational scientists to conduct research. However, most of the existing web crawler frameworks and software products either require professional coding skills without an easy-to-use graphic user interface or are expensive and limited in features. They are thus not friendly to newbies and inconvenient for complicated web-crawling tasks. In this paper, we present an easy-to-use visual web crawler system, EasySpider, for designing and executing web crawling tasks without coding. The workflow of a new web crawling task can be visually programmed by following EasySpider’s visual wizard on the target webpages using an intuitive point-and-click interface. The generated crawler task can then be easily invoked locally or as a web service. Our EasySpider is cross-platform and flexible to adapt to different web-resources. It also supports advanced configuration for complicated tasks and extension. The whole system is open-sourced and transparent for free-access at GitHub, which avoids possible privacy leakage.
We propose a speed-up method for the in-focus plane detection in digital holographic microscopy that can be applied to a broad class of autofocusing algorithms that involve repetitive propagation of an object wave to various axial locations to decide the in-focus position. The classical autofocusing algorithms apply a uniform search strategy, i.e., they probe multiple, uniformly distributed axial locations, which leads to heavy computational overhead. Our method substantially reduces the computational load, without sacrificing the accuracy, by skillfully selecting the next location to investigate, which results in a decreased total number of probed propagation distances. This is achieved by applying the golden selection search with parabolic interpolation, which is the gold standard for tackling single-variable optimization problems. The proposed approach is successfully applied to three diverse autofocusing cases, providing up to 136-fold speed-up.
Traditional supervised learning tasks require a label for every instance in the training set, but in many real-world applications, labels are only available for collections (bags) of instances. This problem setting, known as multiple instance learning (MIL), is particularly relevant in the medical domain, where high-resolution images are split into smaller patches, but labels apply to the image as a whole. Recent MIL models are able to capture correspondences between patches by employing self-attention, allowing them to weigh each patch differently based on all other patches in the bag. However, these approaches still do not consider the relative spatial relationships between patches within the larger image, which is especially important in computational pathology. To this end, we introduce a novel MIL model with distance-aware self-attention (DAS-MIL), which explicitly takes into account relative spatial information when modelling the interactions between patches. Unlike existing relative position representations for self-attention which are discrete, our approach introduces continuous distance-dependent terms into the computation of the attention weights, and is the first to apply relative position representations in the context of MIL. We evaluate our model on a custom MNIST-based MIL dataset that requires the consideration of relative spatial information, as well as on CAMELYON16, a publicly available cancer metastasis detection dataset, where we achieve a test AUROC score of 0.91. On both datasets, our model outperforms existing MIL approaches that employ absolute positional encodings, as well as existing relative position representation schemes applied to MIL. Our code is available at https://anonymous.4open.science/r/das-mil.
Synthesizing visual content that meets users' needs often requires flexible and precise controllability of the pose, shape, expression, and layout of the generated objects. Existing approaches gain controllability of generative adversarial networks (GANs) via manually annotated training data or a prior 3D model, which often lack flexibility, precision, and generality. In this work, we study a powerful yet much less explored way of controlling GANs, that is, to "drag" any points of the image to precisely reach target points in a user-interactive manner, as shown in Fig.1. To achieve this, we propose DragGAN, which consists of two main components: 1) a feature-based motion supervision that drives the handle point to move towards the target position, and 2) a new point tracking approach that leverages the discriminative generator features to keep localizing the position of the handle points. Through DragGAN, anyone can deform an image with precise control over where pixels go, thus manipulating the pose, shape, expression, and layout of diverse categories such as animals, cars, humans, landscapes, etc. As these manipulations are performed on the learned generative image manifold of a GAN, they tend to produce realistic outputs even for challenging scenarios such as hallucinating occluded content and deforming shapes that consistently follow the object's rigidity. Both qualitative and quantitative comparisons demonstrate the advantage of DragGAN over prior approaches in the tasks of image manipulation and point tracking. We also showcase the manipulation of real images through GAN inversion.
Chest X-rays have been widely used for COVID-19 screening; however, 3D computed tomography (CT) is a more effective modality. We present our findings on COVID-19 severity prediction from chest CT scans using the STOIC dataset. We developed an ensemble deep learning based model that incorporates multiple neural networks to improve predictions. To address data imbalance, we used slicing functions and data augmentation. We further improved performance using test time data augmentation. Our approach which employs a simple yet effective ensemble of deep learning-based models with strong test time augmentations, achieved results comparable to more complex methods and secured the fourth position in the STOIC2021 COVID-19 AI Challenge. Our code is available on online: at: https://github.com/aleemsidra/stoic2021- baseline-finalphase-main.
In most Vision-Language models (VL), the understanding of the image structure is enabled by injecting the position information (PI) about objects in the image. In our case study of LXMERT, a state-of-the-art VL model, we probe the use of the PI in the representation and study its effect on Visual Question Answering. We show that the model is not capable of leveraging the PI for the image-text matching task on a challenge set where only position differs. Yet, our experiments with probing confirm that the PI is indeed present in the representation. We introduce two strategies to tackle this: (i) Positional Information Pre-training and (ii) Contrastive Learning on PI using Cross-Modality Matching. Doing so, the model can correctly classify if images with detailed PI statements match. Additionally to the 2D information from bounding boxes, we introduce the object's depth as new feature for a better object localization in the space. Even though we were able to improve the model properties as defined by our probes, it only has a negligible effect on the downstream performance. Our results thus highlight an important issue of multimodal modeling: the mere presence of information detectable by a probing classifier is not a guarantee that the information is available in a cross-modal setup.
In implant prosthesis treatment, the design of the surgical guide heavily relies on the manual location of the implant position, which is subjective and prone to doctor's experiences. When deep learning based methods has started to be applied to address this problem, the space between teeth are various and some of them might present similar texture characteristic with the actual implant region. Both problems make a big challenge for the implant position prediction. In this paper, we develop a two-stream implant position regression framework (TSIPR), which consists of an implant region detector (IRD) and a multi-scale patch embedding regression network (MSPENet), to address this issue. For the training of IRD, we extend the original annotation to provide additional supervisory information, which contains much more rich characteristic and do not introduce extra labeling costs. A multi-scale patch embedding module is designed for the MSPENet to adaptively extract features from the images with various tooth spacing. The global-local feature interaction block is designed to build the encoder of MSPENet, which combines the transformer and convolution for enriched feature representation. During inference, the RoI mask extracted from the IRD is used to refine the prediction results of the MSPENet. Extensive experiments on a dental implant dataset through five-fold cross-validation demonstrated that the proposed TSIPR achieves superior performance than existing methods.
A framework for tightly integrated motion mode classification and state estimation in motion-constrained inertial navigation systems is presented. The framework uses a jump Markov model to describe the navigation system's motion mode and navigation state dynamics with a single model. A bank of Kalman filters is then used for joint inference of the navigation state and the motion mode. A method for learning unknown parameters in the jump Markov model, such as the motion mode transition probabilities, is also presented. The application of the proposed framework is illustrated via two examples. The first example is a foot-mounted navigation system that adapts its behavior to different gait speeds. The second example is a foot-mounted navigation system that detects when the user walks on flat ground and locks the vertical position estimate accordingly. Both examples show that the proposed framework provides significantly better position accuracy than a standard zero-velocity aided inertial navigation system. More importantly, the examples show that the proposed framework provides a theoretically well-grounded approach for developing new motion-constrained inertial navigation systems that can learn different motion patterns.
Automatically narrating videos in natural language complying with user requests, i.e. Controllable Video Captioning task, can help people manage massive videos with desired intentions. However, existing works suffer from two shortcomings: 1) the control signal is single-grained which can not satisfy diverse user intentions; 2) the video description is generated in a single round which can not be further edited to meet dynamic needs. In this paper, we propose a novel \textbf{V}ideo \textbf{C}aption \textbf{E}diting \textbf{(VCE)} task to automatically revise an existing video description guided by multi-grained user requests. Inspired by human writing-revision habits, we design the user command as a pivotal triplet \{\textit{operation, position, attribute}\} to cover diverse user needs from coarse-grained to fine-grained. To facilitate the VCE task, we \textit{automatically} construct an open-domain benchmark dataset named VATEX-EDIT and \textit{manually} collect an e-commerce dataset called EMMAD-EDIT. We further propose a specialized small-scale model (i.e., OPA) compared with two generalist Large Multi-modal Models to perform an exhaustive analysis of the novel task. For evaluation, we adopt comprehensive metrics considering caption fluency, command-caption consistency, and video-caption alignment. Experiments reveal the task challenges of fine-grained multi-modal semantics understanding and processing. Our datasets, codes, and evaluation tools are available at https://github.com/yaolinli/VCE.
In the last five years, there has been a significant focus in Natural Language Processing (NLP) on developing larger Pretrained Language Models (PLMs) and introducing benchmarks such as SuperGLUE and SQuAD to measure their abilities in language understanding, reasoning, and reading comprehension. These PLMs have achieved impressive results on these benchmarks, even surpassing human performance in some cases. This has led to claims of superhuman capabilities and the provocative idea that certain tasks have been solved. In this position paper, we take a critical look at these claims and ask whether PLMs truly have superhuman abilities and what the current benchmarks are really evaluating. We show that these benchmarks have serious limitations affecting the comparison between humans and PLMs and provide recommendations for fairer and more transparent benchmarks.
This paper studies the equilibrium behavior in contests with stochastic progress. Participants have access to a safe action that makes progress deterministically, but they can also take risky moves that stochastically influence their progress towards the goal and thus their relative position. In the unique well-behaved Markov perfect equilibrium of this dynamic contest, the follower drops out if the leader establishes a substantial lead, but resorts to "Hail Marys" beforehand: no matter how low the return of the risky move is, the follower undertakes in an attempt to catch up. Moreover, if the risky move has a medium return (between high and low), the leader will also adopt it when the follower is close to dropping out - an interesting preemptive motive. We also examine the impact of such equilibrium behavior on the optimal prize allocation.
Recently, in speech emotion recognition, a Transformer-based method using spectrogram images instead of sound data showed improved accuracy than Convolutional Neural Networks (CNNs). Vision Transformer (ViT), a Transformer-based method, achieves high classification accuracy by using divided patches from the input image, but has a problem in that pixel position information is not retained due to embedding layers such as linear projection. Therefore, in this paper, we propose a novel method of improve ViT-based speech emotion recognition using coordinate information concatenate. Since the proposed method retains pixel position information by concatenating coordinate information to the input image, the accuracy of CREMA-D is greatly improved by 82.96% compared to the state-of-art about CREMA-D. As a result, it proved that the coordinate information concatenate proposed in this paper is effective not only for CNNs but also for Transformers.
Although Maxwell discovered the physical laws of electromagnetic waves 160 years ago, how to precisely model the propagation of an RF signal in an electrically large and complex environment remains a long-standing problem. The difficulty is in the complex interactions between the RF signal and the obstacles (e.g., reflection, diffraction, etc.). Inspired by the great success of using a neural network to describe the optical field in computer vision, we propose a neural radio-frequency radiance field, NeRF$^\textbf{2}$, which represents a continuous volumetric scene function that makes sense of an RF signal's propagation. Particularly, after training with a few signal measurements, NeRF$^\textbf{2}$ can tell how/what signal is received at any position when it knows the position of a transmitter. As a physical-layer neural network, NeRF$^\textbf{2}$ can take advantage of the learned statistic model plus the physical model of ray tracing to generate a synthetic dataset that meets the training demands of application-layer artificial neural networks (ANNs). Thus, we can boost the performance of ANNs by the proposed turbo-learning, which mixes the true and synthetic datasets to intensify the training. Our experiment results show that turbo-learning can enhance performance with an approximate 50% increase. We also demonstrate the power of NeRF$^\textbf{2}$ in the field of indoor localization and 5G MIMO.
End-to-end autonomous driving has made impressive progress in recent years. Existing methods usually adopt the decoupled encoder-decoder paradigm, where the encoder extracts hidden features from raw sensor data, and the decoder outputs the ego-vehicle's future trajectories or actions. Under such a paradigm, the encoder does not have access to the intended behavior of the ego agent, leaving the burden of finding out safety-critical regions from the massive receptive field and inferring about future situations to the decoder. Even worse, the decoder is usually composed of several simple multi-layer perceptrons (MLP) or GRUs while the encoder is delicately designed (e.g., a combination of heavy ResNets or Transformer). Such an imbalanced resource-task division hampers the learning process. In this work, we aim to alleviate the aforementioned problem by two principles: (1) fully utilizing the capacity of the encoder; (2) increasing the capacity of the decoder. Concretely, we first predict a coarse-grained future position and action based on the encoder features. Then, conditioned on the position and action, the future scene is imagined to check the ramification if we drive accordingly. We also retrieve the encoder features around the predicted coordinate to obtain fine-grained information about the safety-critical region. Finally, based on the predicted future and the retrieved salient feature, we refine the coarse-grained position and action by predicting its offset from ground-truth. The above refinement module could be stacked in a cascaded fashion, which extends the capacity of the decoder with spatial-temporal prior knowledge about the conditioned future. We conduct experiments on the CARLA simulator and achieve state-of-the-art performance in closed-loop benchmarks. Extensive ablation studies demonstrate the effectiveness of each proposed module.
Permanent magnet synchronous motors (PMSM) are widely used due to their numerous benefits. It is critical to get rotor position and speed information in order to operate the motor accurately. Sensorless control techniques have emerged as a popular study area both at home and overseas. The sliding mode observer (SMO) may indirectly detect rotor position and has the benefits of easy implementation and efficient algorithms. In this study, a mathematical model for sensorless control of a PMSM is developed using SMO, vector control, and other techniques. With a surface-mounted PMSM as the study object, a mathematical model for sensorless control of PMSM is developed. PMSM's sliding mode observer model is built in the matlab/simulink environment. Experiments demonstrate that the system can track the rotor position and speed of the motor precisely and fulfill the requirements of sensorless vector control of PMSM.
Generative Adversarial Networks (GANs) have been studied in text generation to tackle the exposure bias problem. Despite their remarkable development, they adopt autoregressive structures so suffering from high latency in both training and inference stages. Although GANs have potential to support efficient generation by adopting non-autoregressive (NAR) structures, their explorations in NAR models are extremely limited. In this work, we conduct pioneering study of building language GANs based on NAR structures. We identify two issues that constrain the performance of GAN-based NAR models. Firstly, existing methods of incorporating latent variables provide highly similar representations which cannot describe the diversity of different words in sentences. We tackle this problem by proposing Position-Aware Self-Modulation, providing more diverse and effective representations. Secondly, the attention mechanism in Transformer cannot accurately build word dependencies in the unstable training of GANs, and we adopt Dependency Feed Forward Network to enhance the model capacity in dependency modeling. Armed with these two facilities, we propose a GAN-based NAR model, Adversarial Non-autoregressive Transformer (ANT). The experimental results demonstrate that ANT can achieve comparable performance with mainstream models in a single forward pass and has great potential in various applications like latent interpolation and semi-supervised learning.
Sequence modeling has important applications in natural language processing and computer vision. Recently, the transformer-based models have shown strong performance on various sequence modeling tasks, which rely on attention to capture pairwise token relations, and position embedding to inject positional information. While showing good performance, the transformer models are inefficient to scale to long input sequences, mainly due to the quadratic space-time complexity of attention. To overcome this inefficiency, we propose to model sequences with a relative position encoded Toeplitz matrix and use a Toeplitz matrix-vector production trick to reduce the space-time complexity of the sequence modeling to log linear. A lightweight sub-network called relative position encoder is proposed to generate relative position coefficients with a fixed budget of parameters, enabling the proposed Toeplitz neural network to deal with varying sequence lengths. In addition, despite being trained on 512-token sequences, our model can extrapolate input sequence length up to 14K tokens in inference with consistent performance. Extensive experiments on autoregressive and bidirectional language modeling, image modeling, and the challenging Long-Range Arena benchmark show that our method achieves better performance than its competitors in most downstream tasks while being significantly faster. The code is available at https://github.com/OpenNLPLab/Tnn.