arXiv subjects

Chang Xu

Publications and source records attributed to Chang Xu.

At least 271 records · Page 15Linked to original sources

arXiv2022

LightViT: Towards Light-Weight Convolution-Free Vision Transformers

Vision transformers (ViTs) are usually considered to be less light-weight than convolutional neural networks (CNNs) due to the lack of inductive bias. Recent works thus resort to convolutions as a plug-and-play module and embed them in various ViT counterparts. In this paper, we argue that the convolutional kernels perform information aggregation to connect all tokens; however, they would be actually unnecessary for light-weight ViTs if this explicit aggregation could function in a more homogeneous way. Inspired by this, we present LightViT as a new family of light-weight ViTs to achieve better accuracy-efficiency balance upon the pure transformer blocks without convolution. Concretely, we introduce a global yet efficient aggregation scheme into both self-attention and feed-forward network (FFN) of ViTs, where additional learnable tokens are introduced to capture global dependencies; and bi-dimensional channel and spatial attentions are imposed over token embeddings. Experiments show that our model achieves significant improvements on image classification, object detection, and semantic segmentation tasks. For example, our LightViT-T achieves 78.7% accuracy on ImageNet with only 0.7G FLOPs, outperforming PVTv2-B0 by 8.2% while 11% faster on GPU. Code is available at https://github.com/hunto/LightViT.

cs.CV↗

arXiv2022

Detecting tiny objects in aerial images: A normalized Wasserstein distance and a new benchmark

Tiny object detection (TOD) in aerial images is challenging since a tiny object only contains a few pixels. State-of-the-art object detectors do not provide satisfactory results on tiny objects due to the lack of supervision from discriminative features. Our key observation is that the Intersection over Union (IoU) metric and its extensions are very sensitive to the location deviation of the tiny objects, which drastically deteriorates the quality of label assignment when used in anchor-based detectors. To tackle this problem, we propose a new evaluation metric dubbed Normalized Wasserstein Distance (NWD) and a new RanKing-based Assigning (RKA) strategy for tiny object detection. The proposed NWD-RKA strategy can be easily embedded into all kinds of anchor-based detectors to replace the standard IoU threshold-based one, significantly improving label assignment and providing sufficient supervision information for network training. Tested on four datasets, NWD-RKA can consistently improve tiny object detection performance by a large margin. Besides, observing prominent noisy labels in the Tiny Object Detection in Aerial Images (AI-TOD) dataset, we are motivated to meticulously relabel it and release AI-TOD-v2 and its corresponding benchmark. In AI-TOD-v2, the missing annotation and location error problems are considerably mitigated, facilitating more reliable training and validation processes. Embedding NWD-RKA into DetectoRS, the detection performance achieves 4.3 AP points improvement over state-of-the-art competitors on AI-TOD-v2. Datasets, codes, and more visualizations are available at: https://chasel-tsui.github.io/AI-TOD-v2/

cs.CV↗

arXiv2022

A $q$-supercongruence modulo the fourth power of a cyclotomic polynomial

In this paper, a new $q$-supercongruence with two free parameters modulo the fourth power of a cyclotomic polynomial is obtained. Our main auxiliary tools are Watson's $_8ϕ_7$ transformation formula for basic hypergeometric series, the `creative microscoping' method recently introduced by Guo and Zudilin and the Chinese remainder theorem for coprime polynomials. By taking suitable parameter substitutions in the established $q$-supercongruence, some nice congruences involving the Bernoulli numbers are derived.

math.NT↗

arXiv2022

CMT: Convolutional Neural Networks Meet Vision Transformers

Vision transformers have been successfully applied to image recognition tasks due to their ability to capture long-range dependencies within an image. However, there are still gaps in both performance and computational cost between transformers and existing convolutional neural networks (CNNs). In this paper, we aim to address this issue and develop a network that can outperform not only the canonical transformers, but also the high-performance convolutional models. We propose a new transformer based hybrid network by taking advantage of transformers to capture long-range dependencies, and of CNNs to model local features. Furthermore, we scale it to obtain a family of models, called CMTs, obtaining much better accuracy and efficiency than previous convolution and transformer based models. In particular, our CMT-S achieves 83.5% top-1 accuracy on ImageNet, while being 14x and 2x smaller on FLOPs than the existing DeiT and EfficientNet, respectively. The proposed CMT-S also generalizes well on CIFAR10 (99.2%), CIFAR100 (91.7%), Flowers (98.7%), and other challenging vision datasets such as COCO (44.3% mAP), with considerably less computational cost.

cs.CV↗

arXiv2022

A Normalized Gaussian Wasserstein Distance for Tiny Object Detection

Detecting tiny objects is a very challenging problem since a tiny object only contains a few pixels in size. We demonstrate that state-of-the-art detectors do not produce satisfactory results on tiny objects due to the lack of appearance information. Our key observation is that Intersection over Union (IoU) based metrics such as IoU itself and its extensions are very sensitive to the location deviation of the tiny objects, and drastically deteriorate the detection performance when used in anchor-based detectors. To alleviate this, we propose a new evaluation metric using Wasserstein distance for tiny object detection. Specifically, we first model the bounding boxes as 2D Gaussian distributions and then propose a new metric dubbed Normalized Wasserstein Distance (NWD) to compute the similarity between them by their corresponding Gaussian distributions. The proposed NWD metric can be easily embedded into the assignment, non-maximum suppression, and loss function of any anchor-based detector to replace the commonly used IoU metric. We evaluate our metric on a new dataset for tiny object detection (AI-TOD) in which the average object size is much smaller than existing object detection datasets. Extensive experiments show that, when equipped with NWD metric, our approach yields performance that is 6.7 AP points higher than a standard fine-tuning baseline, and 6.0 AP points higher than state-of-the-art competitors. Codes are available at: https://github.com/jwwangchn/NWD.

cs.CV↗

arXiv2022

Heavy + heavy and heavy + light pseudoscalar to vector semileptonic transitions

Using a symmetry-preserving regularisation of a vector$\times$vector contact interaction (SCI), we complete a systematic treatment of twelve semileptonic transitions with vector meson final states: $D\to ρ$, $D_{(s)}\to K^\ast$, $D_s\to ϕ$, $B\to ρ$, $B_s\to K^\ast$, $B_{(s)}\to D_{(s)}^\ast$, $B_c \to B_{(s)}^\ast, J/ψ, D^\ast$; and thereby finalise a unified analysis of semileptonic decays of heavy+heavy and heavy+light pseudoscalar mesons to both pseudoscalar and vector meson final states. The analysis is marked by algebraic simplicity, few parameters, and the ability to consistently describe systems from Nambu-Goldstone modes to heavy+heavy mesons. Regarding the behaviour of the transition form factors, the SCI results compare well wherever sound experimental or independent theory analyses are available; hence, the SCI branching fraction predictions should be a reasonable guide. Considering the ratios $R(D_{(s)}^{(\ast)})$, $R(J/ψ)$, $R(η_c)$, whose values are key tests of lepton universality in weak interactions, the SCI values agree with Standard Model predictions. The $B_{(s)}\to D_{(s)}^\ast$ transitions are used to predict the precursor functions that evolve into the universal Isgur-Wise function in the heavy-quark limit, with results that conform with those from other sources where such are available. The study also exposes effects on the transition form factors that flow from interference between emergent hadron mass from the strong interaction and Higgs boson couplings via current-quark masses, including flavour symmetry violation.

hep-ph↗

arXiv2022

Inertia of two-qutrit entanglement witnesses

Entanglement witnesses (EWs) are a fundamental tool for the detection of entanglement. We investigate the inertias of bipartite EWs constructed by the partial transpose of NPT states. Furthermore, we find out most of the inertias of the patial transpose of the two-qutrit bipartite NPT states. As an application, we extend our results to high dimensional states.

quant-ph↗

arXiv2022

GreedyNASv2: Greedier Search with a Greedy Path Filter

Training a good supernet in one-shot NAS methods is difficult since the search space is usually considerably huge (e.g., $13^{21}$). In order to enhance the supernet's evaluation ability, one greedy strategy is to sample good paths, and let the supernet lean towards the good ones and ease its evaluation burden as a result. However, in practice the search can be still quite inefficient since the identification of good paths is not accurate enough and sampled paths still scatter around the whole search space. In this paper, we leverage an explicit path filter to capture the characteristics of paths and directly filter those weak ones, so that the search can be thus implemented on the shrunk space more greedily and efficiently. Concretely, based on the fact that good paths are much less than the weak ones in the space, we argue that the label of "weak paths" will be more confident and reliable than that of "good paths" in multi-path sampling. In this way, we thus cast the training of path filter in the positive and unlabeled (PU) learning paradigm, and also encourage a \textit{path embedding} as better path/operation representation to enhance the identification capacity of the learned filter. By dint of this embedding, we can further shrink the search space by aggregating similar operations with similar embeddings, and the search can be more efficient and accurate. Extensive experiments validate the effectiveness of the proposed method GreedyNASv2. For example, our obtained GreedyNASv2-L achieves $81.1\%$ Top-1 accuracy on ImageNet dataset, significantly outperforming the ResNet-50 strong baselines.

cs.CV↗

arXiv2022

3D Visual Tracking to Quantify Physical Contact Interactions in Human-to-Human Touch

Across a plethora of social situations, we touch others in natural and intuitive ways to share thoughts and emotions, such as tapping to get one's attention or caressing to soothe one's anxiety. A deeper understanding of these human-to-human interactions will require, in part, the precise measurement of skin-to-skin physical contact. Among prior efforts, each measurement approach exhibits certain constraints, e.g., motion trackers do not capture the precise shape of skin surfaces, while pressure sensors impede direct skin contact. In contrast, this work develops an interference-free 3D visual tracking system using a depth camera to measure the contact attributes between the bare hand of a toucher and the forearm of a receiver. The toucher's hand is tracked as a posed and positioned mesh by fitting a hand model to detected 3D joints, whereas the forearm is extracted as a detailed 3D surface. Based on a contact model of point clouds, the spatiotemporal contact changes are decomposed as six high-resolution time-series attributes, i.e., contact area, indentation depth, absolute velocity, and three orthogonal velocity components, together with contact duration. To examine the system's capabilities and limitations, two experiments were performed. First, to evaluate its ability to discern human touches, one person delivered cued social messages, e.g., happiness, anger, sympathy, to another using their preferred gestures. The results indicated that messages, gestures, and the individual touchers, were readily discerned from their contact attributes. Second, the measurement accuracy was validated against independent devices, including an electromagnetic tracker, pressure mat, and laser sensor. While validated here in the context of social communication, this system is extendable to human touch interactions such as maternal care of infants and massage therapy.

cs.HC↗

arXiv2022

An Image Patch is a Wave: Phase-Aware Vision MLP

In the field of computer vision, recent works show that a pure MLP architecture mainly stacked by fully-connected layers can achieve competing performance with CNN and transformer. An input image of vision MLP is usually split into multiple tokens (patches), while the existing MLP models directly aggregate them with fixed weights, neglecting the varying semantic information of tokens from different images. To dynamically aggregate tokens, we propose to represent each token as a wave function with two parts, amplitude and phase. Amplitude is the original feature and the phase term is a complex value changing according to the semantic contents of input images. Introducing the phase term can dynamically modulate the relationship between tokens and fixed weights in MLP. Based on the wave-like token representation, we establish a novel Wave-MLP architecture for vision tasks. Extensive experiments demonstrate that the proposed Wave-MLP is superior to the state-of-the-art MLP architectures on various vision tasks such as image classification, object detection and semantic segmentation. The source code is available at https://github.com/huawei-noah/CV-Backbones/tree/master/wavemlp_pytorch and https://gitee.com/mindspore/models/tree/master/research/cv/wave_mlp.

cs.CV↗

arXiv2022

Patch Slimming for Efficient Vision Transformers

This paper studies the efficiency problem for visual transformers by excavating redundant calculation in given networks. The recent transformer architecture has demonstrated its effectiveness for achieving excellent performance on a series of computer vision tasks. However, similar to that of convolutional neural networks, the huge computational cost of vision transformers is still a severe issue. Considering that the attention mechanism aggregates different patches layer-by-layer, we present a novel patch slimming approach that discards useless patches in a top-down paradigm. We first identify the effective patches in the last layer and then use them to guide the patch selection process of previous layers. For each layer, the impact of a patch on the final output feature is approximated and patches with less impact will be removed. Experimental results on benchmark datasets demonstrate that the proposed method can significantly reduce the computational costs of vision transformers without affecting their performances. For example, over 45% FLOPs of the ViT-Ti model can be reduced with only 0.2% top-1 accuracy drop on the ImageNet dataset.

cs.CV↗

arXiv2022

Strain-dependent structural and electronic reconstructions in long-wavelength WS$_{2}$ moiré superlattices

In long-wavelength moiré superlattices of stacked transition metal dichalcogenides (TMDs), structural reconstruction ubiquitously occurs, which has reported to impact significantly their electronic properties. However, complete microscopic understandings of the interplay between the lattice reconstruction and alteration of electronic properties, and their further response to external perturbations in the reconstructed TMDs moiré superlattice are still lacking. Here, using scanning tunneling microscopy (STM) and scanning tunneling spectroscopy (STS) combined with first-principles calculation, we study the strain-dependent structural reconstruction and its correlated electronic reconstruction in long-wavelength H-type WS$_{2}$ moiré superlattice at nanometer scale. We observe that the long-wavelength WS$_{2}$ moiré superlattices experiencing strong atomic reconstruction transform into a hexagonal array of screw dislocations separating large-sized H-stacked domains. Both the geometry and the moiré wavelength of the moiré superlattice are dramatically tuned by external intralayer heterostrain in our experiment. Remarkably, the STS measurements further demonstrate that the location of the K point in conduction band is modulated sensitively by strain-induced lattice deformation at nanometer scale in this system, with the maximum energy shift reaching up to 300 meV. Our results highlight that intralayer strain plays a vital role in determining structural and electronic properties in TMD moiré superlattice.

cond-mat.mes-hall↗

arXiv2022

Searching for Network Width with Bilaterally Coupled Network

Searching for a more compact network width recently serves as an effective way of channel pruning for the deployment of convolutional neural networks (CNNs) under hardware constraints. To fulfill the searching, a one-shot supernet is usually leveraged to efficiently evaluate the performance \wrt~different network widths. However, current methods mainly follow a \textit{unilaterally augmented} (UA) principle for the evaluation of each width, which induces the training unfairness of channels in supernet. In this paper, we introduce a new supernet called Bilaterally Coupled Network (BCNet) to address this issue. In BCNet, each channel is fairly trained and responsible for the same amount of network widths, thus each network width can be evaluated more accurately. Besides, we propose to reduce the redundant search space and present the BCNetV2 as the enhanced supernet to ensure rigorous training fairness over channels. Furthermore, we leverage a stochastic complementary strategy for training the BCNet, and propose a prior initial population sampling method to boost the performance of the evolutionary search. We also propose the first open-source width benchmark on macro structures named Channel-Bench-Macro for the better comparison of width search algorithms. Extensive experiments on benchmark CIFAR-10 and ImageNet datasets indicate that our method can achieve state-of-the-art or competing performance over other baseline methods. Moreover, our method turns out to further boost the performance of NAS models by refining their network widths. For example, with the same FLOPs budget, our obtained EfficientNet-B0 achieves 77.53\% Top-1 accuracy on ImageNet dataset, surpassing the performance of original setting by 0.65\%.

cs.CV↗

arXiv2022

DyRep: Bootstrapping Training with Dynamic Re-parameterization

Structural re-parameterization (Rep) methods achieve noticeable improvements on simple VGG-style networks. Despite the prevalence, current Rep methods simply re-parameterize all operations into an augmented network, including those that rarely contribute to the model's performance. As such, the price to pay is an expensive computational overhead to manipulate these unnecessary behaviors. To eliminate the above caveats, we aim to bootstrap the training with minimal cost by devising a dynamic re-parameterization (DyRep) method, which encodes Rep technique into the training process that dynamically evolves the network structures. Concretely, our proposal adaptively finds the operations which contribute most to the loss in the network, and applies Rep to enhance their representational capacity. Besides, to suppress the noisy and redundant operations introduced by Rep, we devise a de-parameterization technique for a more compact re-parameterization. With this regard, DyRep is more efficient than Rep since it smoothly evolves the given network instead of constructing an over-parameterized network. Experimental results demonstrate our effectiveness, e.g., DyRep improves the accuracy of ResNet-18 by $2.04\%$ on ImageNet and reduces $22\%$ runtime over the baseline. Code is available at: https://github.com/hunto/DyRep.

cs.CV↗

arXiv2022

$q$-Supercongruences on triple and quadruple sums

Inspired by the recent work of El Bachraoui, we present some new $q$-supercongruences on triple and quadruple sums of basic hypergeometric series. In particular, we give a $q$-supercongruence modulo the fifth power of a cyclotomic polynomial, which is a $q$-analogue of the quadruple sum of Van Hamme's supercongruence (G.2).

math.NT↗

arXiv2022

SimMatch: Semi-supervised Learning with Similarity Matching

Learning with few labeled data has been a longstanding problem in the computer vision and machine learning research community. In this paper, we introduced a new semi-supervised learning framework, SimMatch, which simultaneously considers semantic similarity and instance similarity. In SimMatch, the consistency regularization will be applied on both semantic-level and instance-level. The different augmented views of the same instance are encouraged to have the same class prediction and similar similarity relationship respected to other instances. Next, we instantiated a labeled memory buffer to fully leverage the ground truth labels on instance-level and bridge the gaps between the semantic and instance similarities. Finally, we proposed the \textit{unfolding} and \textit{aggregation} operation which allows these two similarities be isomorphically transformed with each other. In this way, the semantic and instance pseudo-labels can be mutually propagated to generate more high-quality and reliable matching targets. Extensive experimental results demonstrate that SimMatch improves the performance of semi-supervised learning tasks across different benchmark datasets and different settings. Notably, with 400 epochs of training, SimMatch achieves 67.2\%, and 74.4\% Top-1 Accuracy with 1\% and 10\% labeled examples on ImageNet, which significantly outperforms the baseline methods and is better than previous semi-supervised learning frameworks. Code and pre-trained models are available at https://github.com/KyleZheng1997/simmatch.

cs.CV↗

arXiv2022

Some new results about $q$-trinomial coefficients

In this paper, we present several new congruences on the $q$-trinomial coefficients introduced by Andrews and Baxter. A new congruence on sums of central $q$-binomial coefficients is also established.

math.NT↗

arXiv2022

Relational Surrogate Loss Learning

Evaluation metrics in machine learning are often hardly taken as loss functions, as they could be non-differentiable and non-decomposable, e.g., average precision and F1 score. This paper aims to address this problem by revisiting the surrogate loss learning, where a deep neural network is employed to approximate the evaluation metrics. Instead of pursuing an exact recovery of the evaluation metric through a deep neural network, we are reminded of the purpose of the existence of these evaluation metrics, which is to distinguish whether one model is better or worse than another. In this paper, we show that directly maintaining the relation of models between surrogate losses and metrics suffices, and propose a rank correlation-based optimization method to maximize this relation and learn surrogate losses. Compared to previous works, our method is much easier to optimize and enjoys significant efficiency and performance gains. Extensive experiments show that our method achieves improvements on various tasks including image classification and neural machine translation, and even outperforms state-of-the-art methods on human pose estimation and machine reading comprehension tasks. Code is available at: https://github.com/hunto/ReLoss.

cs.LG↗