Object tracking has achieved significant progress over the past few years. However, state-of-the-art trackers become increasingly heavy and expensive, which limits their deployments in resource-constrained applications. In this work, we present LightTrack, which uses neural architecture search (NAS) to design more lightweight and efficient object trackers. Comprehensive experiments show that our LightTrack is effective. It can find trackers that achieve superior performance compared to handcrafted SOTA trackers, such as SiamRPN++ and Ocean, while using much fewer model Flops and parameters. Moreover, when deployed on resource-constrained mobile chipsets, the discovered trackers run much faster. For example, on Snapdragon 845 Adreno GPU, LightTrack runs $12\times$ faster than Ocean, while using $13\times$ fewer parameters and $38\times$ fewer Flops. Such improvements might narrow the gap between academic models and industrial deployments in object tracking task. LightTrack is released at https://github.com/researchmm/LightTrack.
In this paper, we present a new tracking architecture with an encoder-decoder transformer as the key component. The encoder models the global spatio-temporal feature dependencies between target objects and search regions, while the decoder learns a query embedding to predict the spatial positions of the target objects. Our method casts object tracking as a direct bounding box prediction problem, without using any proposals or predefined anchors. With the encoder-decoder transformer, the prediction of objects just uses a simple fully-convolutional network, which estimates the corners of objects directly. The whole method is end-to-end, does not need any postprocessing steps such as cosine window and bounding box smoothing, thus largely simplifying existing tracking pipelines. The proposed tracker achieves state-of-the-art performance on five challenging short-term and long-term benchmarks, while running at real-time speed, being 6x faster than Siam R-CNN. Code and models are open-sourced at https://github.com/researchmm/Stark.
Correlation acts as a critical role in the tracking field, especially in recent popular Siamese-based trackers. The correlation operation is a simple fusion manner to consider the similarity between the template and the search region. However, the correlation operation itself is a local linear matching process, leading to lose semantic information and fall into local optimum easily, which may be the bottleneck of designing high-accuracy tracking algorithms. Is there any better feature fusion method than correlation? To address this issue, inspired by Transformer, this work presents a novel attention-based feature fusion network, which effectively combines the template and search region features solely using attention. Specifically, the proposed method includes an ego-context augment module based on self-attention and a cross-feature augment module based on cross-attention. Finally, we present a Transformer tracking (named TransT) method based on the Siamese-like feature extraction backbone, the designed attention-based fusion mechanism, and the classification and regression head. Experiments show that our TransT achieves very promising results on six challenging datasets, especially on large-scale LaSOT, TrackingNet, and GOT-10k benchmarks. Our tracker runs at approximatively 50 fps on GPU. Code and models are available at https://github.com/chenxin-dlut/TransT.
Visual object tracking aims to precisely estimate the bounding box for the given target, which is a challenging problem due to factors such as deformation and occlusion. Many recent trackers adopt the multiple-stage tracking strategy to improve the quality of bounding box estimation. These methods first coarsely locate the target and then refine the initial prediction in the following stages. However, existing approaches still suffer from limited precision, and the coupling of different stages severely restricts the method's transferability. This work proposes a novel, flexible, and accurate refinement module called Alpha-Refine, which can significantly improve the base trackers' prediction quality. By exploring a series of design options, we conclude that the key to successful refinement is extracting and maintaining detailed spatial information as much as possible. Following this principle, Alpha-Refine adopts a pixel-wise correlation, a corner prediction head, and an auxiliary mask head as the core components. We apply Alpha-Refine to six famous base trackers to verify our method's effectiveness: DiMPsuper, DiMP50, ATOM, SiamRPN++, RT-MDNet, and ECO. Comprehensive experiments on TrackingNet, LaSOT, GOT-10K, and VOT2020 benchmarks show that our approach significantly improves the base tracker's performance with little extra latency. Code and pretrained model is available at https://github.com/MasterBin-IIAU/AlphaRefine.
On basis of functional magnetic resonance imaging (fMRI), researchers are devoted to designing visual encoding models to predict the neuron activity of human in response to presented image stimuli and analyze inner mechanism of human visual cortices. Deep network structure composed of hierarchical processing layers forms deep network models by learning features of data on specific task through big dataset. Deep network models have powerful and hierarchical representation of data, and have brought about breakthroughs for visual encoding, while revealing hierarchical structural similarity with the manner of information processing in human visual cortices. However, previous studies almost used image features of those deep network models pre-trained on classification task to construct visual encoding models. Except for deep network structure, the task or corresponding big dataset is also important for deep network models, but neglected by previous studies. Because image classification is a relatively fundamental task, it is difficult to guide deep network models to master high-level semantic representations of data, which causes into that encoding performance for high-level visual cortices is limited. In this study, we introduced one higher-level vision task: image caption (IC) task and proposed the visual encoding model based on IC features (ICFVEM) to encode voxels of high-level visual cortices. Experiment demonstrated that ICFVEM obtained better encoding performance than previous deep network models pre-trained on classification task. In addition, the interpretation of voxels was realized to explore the detailed characteristics of voxels based on the visualization of semantic words, and comparative analysis implied that high-level visual cortices behaved the correlative representation of image content.
Adversarial attack of CNN aims at deceiving models to misbehave by adding imperceptible perturbations to images. This feature facilitates to understand neural networks deeply and to improve the robustness of deep learning models. Although several works have focused on attacking image classifiers and object detectors, an effective and efficient method for attacking single object trackers of any target in a model-free way remains lacking. In this paper, a cooling-shrinking attack method is proposed to deceive state-of-the-art SiameseRPN-based trackers. An effective and efficient perturbation generator is trained with a carefully designed adversarial loss, which can simultaneously cool hot regions where the target exists on the heatmaps and force the predicted bounding box to shrink, making the tracked target invisible to trackers. Numerous experiments on OTB100, VOT2018, and LaSOT datasets show that our method can effectively fool the state-of-the-art SiameseRPN++ tracker by adding small perturbations to the template or the search regions. Besides, our method has good transferability and is able to deceive other top-performance trackers such as DaSiamRPN, DaSiamRPN-UpdateNet, and DiMP. The source codes are available at https://github.com/MasterBin-IIAU/CSA.
In the visual decoding domain, visually reconstructing presented images given the corresponding human brain activity monitored by functional magnetic resonance imaging (fMRI) is difficult, especially when reconstructing viewed natural images. Visual reconstruction is a conditional image generation on fMRI data and thus generative adversarial network (GAN) for natural image generation is recently introduced for this task. Although GAN-based methods have greatly improved, the fidelity and naturalness of reconstruction are still unsatisfactory due to the small number of fMRI data samples and the instability of GAN training. In this study, we proposed a new GAN-based Bayesian visual reconstruction method (GAN-BVRM) that includes a classifier to decode categories from fMRI data, a pre-trained conditional generator to generate natural images of specified categories, and a set of encoding models and evaluator to evaluate generated images. GAN-BVRM employs the pre-trained generator of the prevailing BigGAN to generate masses of natural images, and selects the images that best matches with the corresponding brain activity through the encoding models as the reconstruction of the image stimuli. In this process, the semantic and detailed contents of reconstruction are controlled by decoded categories and encoding models, respectively. GAN-BVRM used the Bayesian manner to avoid contradiction between naturalness and fidelity from current GAN-based methods and thus can improve the advantages of GAN. Experimental results revealed that GAN-BVRM improves the fidelity and naturalness, that is, the reconstruction is natural and similar to the presented image stimuli.
Compared with traditional machine learning models, deep neural networks perform better, especially in image classification tasks. However, they are vulnerable to adversarial examples. Adding small perturbations on examples causes a good-performance model to misclassify the crafted examples, without category differences in the human eyes, and fools deep models successfully. There are two requirements for generating adversarial examples: the attack success rate and image fidelity metrics. Generally, perturbations are increased to ensure the adversarial examples' high attack success rate; however, the adversarial examples obtained have poor concealment. To alleviate the tradeoff between the attack success rate and image fidelity, we propose a method named AdvJND, adding visual model coefficients, just noticeable difference coefficients, in the constraint of a distortion function when generating adversarial examples. In fact, the visual subjective feeling of the human eyes is added as a priori information, which decides the distribution of perturbations, to improve the image quality of adversarial examples. We tested our method on the FashionMNIST, CIFAR10, and MiniImageNet datasets. Adversarial examples generated by our AdvJND algorithm yield gradient distributions that are similar to those of the original inputs. Hence, the crafted noise can be hidden in the original inputs, thus improving the attack concealment significantly.
Adversarial examples reveal the vulnerability and unexplained nature of neural networks. Studying the defense of adversarial examples is of considerable practical importance. Most adversarial examples that misclassify networks are often undetectable by humans. In this paper, we propose a defense model to train the classifier into a human-perception classification model with shape preference. The proposed model comprising a texture transfer network (TTN) and an auxiliary defense generative adversarial networks (GAN) is called Human-perception Auxiliary Defense GAN (HAD-GAN). The TTN is used to extend the texture samples of a clean image and helps classifiers focus on its shape. GAN is utilized to form a training framework for the model and generate the necessary images. A series of experiments conducted on MNIST, Fashion-MNIST and CIFAR10 show that the proposed model outperforms the state-of-the-art defense methods for network robustness. The model also demonstrates a significant improvement on defense capability of adversarial examples.