FCFei Cheng

Fei Cheng

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Fei Cheng has a total of 27 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    LightCL: Compact Continual Learning with Low Memory Footprint For Edge Device

    by Zeqing Wang, Fei Cheng, Kangye Ji, Bohu Huang

    Continual learning (CL) is a technique that enables neural networks to constantly adapt to their dynamic surroundings. Despite being overlooked for a long time, this technology can considerably address the customized needs of users in edge devices. Actually, most CL methods require huge resource consumption by the training behavior to acquire generalizability among all tasks for delaying forgetting regardless of edge scenarios. Therefore, this paper proposes a compact algorithm called LightCL, which evaluates and compresses the redundancy of already generalized components in structures of the neural network. Specifically, we consider two factors of generalizability, learning plasticity and memory stability, and design metrics of both to quantitatively assess generalizability of neural networks during CL. This evaluation shows that generalizability of different layers in a neural network exhibits a significant variation. Thus, we Maintain Generalizability by freezing generalized parts without the resource-intensive training process and Memorize Feature Patterns by stabilizing feature extracting of previous tasks to enhance generalizability for less-generalized parts with a little extra memory, which is far less than the reduction by freezing. Experiments illustrate that LightCL outperforms other state-of-the-art methods and reduces at most 6.16× memory footprint. We also verify the effectiveness of LightCL on the edge device.


  • Open Access

    EnNet: Enhanced Interactive Information Network with Zero-Order Optimization

    by Yingzhao Shao, Yanxin Chen, Pengfei Yang, Fei Cheng

    Interactive image segmentation extremely accelerates the generation of high-quality annotation image datasets, which are the pillars of the applications of deep learning. However, these methods suffer from the insignificance of interaction information and excessively high optimization costs, resulting in unexpected segmentation outcomes and increased computational burden. To address these issues, this paper focuses on interactive information mining from the network architecture and optimization procedure. In terms of network architecture, the issue mentioned above arises from two perspectives: the less representative feature of interactive regions in each layer and the interactive information weakened by the network hierarchy structure. Therefore, the paper proposes a network called EnNet. The network addresses the two aforementioned issues by employing attention mechanisms to integrate user interaction information across the entire image and incorporating interaction information twice in a design that progresses from coarse to fine. In terms of optimization, this paper proposes a method of using zero-order optimization during the first four iterations of training. This approach can reduce computational overhead with only a minimal reduction in accuracy. The experimental results on GrabCut, Berkeley, DAVIS, and SBD datasets validate the effectiveness of the proposed method, with our approach achieving an average NOC@90 that surpasses RITM by 0.35.

  • Open Access

    JointNet: Multitask Learning Framework for Denoising and Detecting Anomalies in Hyperspectral Remote Sensing

    by Yingzhao Shao, Shuhan Li, Pengfei Yang, Fei Cheng, Yueli Ding, Jianguo Sun

    One of the significant challenges with traditional single-task learning-based anomaly detection using noisy hyperspectral images (HSIs) is the loss of anomaly targets during denoising, especially when the noise and anomaly targets are similar. This issue significantly affects the detection accuracy. To address this problem, this paper proposes a multitask learning (MTL)-based method for detecting anomalies in noisy HSIs. Firstly, a preliminary detection approach based on the JointNet model, which decomposes the noisy HSI into a pure background and a noise–anomaly target mixing component, is introduced. This approach integrates the minimum noise fraction rotation (MNF) algorithm into an autoencoder (AE), effectively isolating the noise while retaining critical features for anomaly detection. Building upon this, the JointNet model is further optimized to ensure that the noise information is shared between the denoising and anomaly detection subtasks, preserving the integrity of the training data during the anomaly detection process and resolving the issue of losing anomaly targets during denoising. A novel loss function is designed to enable the joint learning of both subtasks under the multitask learning model. In addition, a noise score evaluation metric is introduced to calculate the probability of a pixel being an anomaly target, allowing for a clear distinction between noise and anomaly targets, thus providing the final anomaly detection results. The effectiveness of the proposed model and method is validated via testing on the HYDICE and San Diego datasets. The denoising metric results of the PSNR, SSIM, and SAM are 41.79, 0.91, and 4.350 and 42.83, 0.93, and 3.558 on the HYDICE and San Diego datasets, respectively. The anomaly detection ACU is 0.943 and 0.959, respectively. The proposed method outperforms the other algorithms, demonstrating that the reconstructed images using this method exhibited lower noise levels and more complete image information, and the JointNet model outperforms the mainstream HSI anomaly detection algorithms in both the quantitative evaluation and visual effect, showcasing its improved detection capabilities.


  • Fine-grained complexity-driven latency predictor in hardware-aware neural architecture search using composite loss

    by Chengmin Lin, Pengfei Yang, Chengcheng Li, Fei Cheng, Wenkai Lv, Zhenyi Wang, Quan Wang


  • QANS: Toward Quantized Neural Network Adversarial Noise Suppression

    by Chengmin Lin, Fei Cheng, Pengfei Yang, Tianbing He, Jinpeng Liang, Quan Wang

    Neural network quantization techniques play an important role in efficiently deploying deep learning models on the hardware with limited computing and storage resources. Numerous applications of this technology, such as autopilot, necessitate not just efficiency, but also robustness. Research on the robustness of quantized networks against adversarial attacks is becoming one of the major points of interest. In this work, we rethink the impact of quantization on adversarial attacks and explore the boundary of the robustness of quantized neural networks. This study reveals that activation quantization can be used as a defense to weaken adversarial noise, but the robustness of quantized models is still limited by the amplification effect of network errors, including quantization errors and adversarial noise. To address this problem, we propose the Quantization Adversarial Noise Suppression (QANS) method that employs a Gaussian kernel regularization constraint to stabilize the model by restricting the perturbation error within two levels of tolerance. Extensive experiments are conducted with Wide ResNet and VGG-16 models on CIFAR-10 and SVHN datasets under different attack methods, including several white-box and black-box attacks. Experimental results show the proposed method achieves superior robustness to prior works.


  • Spatio-temporal attention on manifold space for 3D human action recognition

    by Chongyang Ding, Kai Liu, Fei Cheng, Evgeny Belyaev


  • Open Access

    Nonnegative Tensor‐Based Linear Dynamical Systems for Recognizing Human Action from 3D Skeletons

    by Guang Li, Kai Liu, Wenwen Ding, Fei Cheng, Chongyang Ding

    Recently, skeleton‐based action recognition has become a very important topic in the field of computer vision. It is a challenging task to accurately build a human action model and precisely distinguish similar human actions. In this paper, an action (skeleton sequence) is represented as a third‐order nonnegative tensor time series to capture the original spatiotemporal information of the action. As a linear dynamical system (LDS) is an efficient tool for encoding the spatiotemporal data in various disciplines, this paper proposes a nonnegative tensor‐based LDS (nLDS) to model the third‐order nonnegative tensor time series. Nonnegative Tucker decomposition (NTD) is utilized to estimate the parameters of the nLDS model. These parameters are used to build extended observability sequence for the action, which implies that can be considered as the feature descriptor of the action. To avoid the limitations introduced by approximating with a finite‐order matrix, we represent an action as a point on infinite Grassmann manifold comprising the orthonormalized extended observability sequences. The classification task can be performed by dictionary learning and sparse coding on the infinite Grassmann manifold. The experimental results on the MSR‐Action3D, UTKinect‐Action, and G3D‐Gaming datasets demonstrate that the proposed approach achieves a better performance in comparison with the state‐of‐the‐art methods.


  • Open Access

    Key-Skeleton-Pattern Mining on 3D Skeletons Represented by Lie Group for Action Recognition

    by Guang Li, Kai Liu, Wenwen Ding, Fei Cheng, Boyang Chen

    The human skeleton can be considered as a tree system of rigid bodies connected by bone joints. In recent researches, substantial progress has been made in both theories and experiments on skeleton-based action recognition. However, it is challenging to accurately represent the skeleton and precisely eliminate noisy skeletons from the action sequence. This paper proposes a novel skeletal representation, which is composed of two subfeatures to recognize human action: static features and dynamic features. First, to avoid scale variations from subject to subject, the orientations of the rigid bodies in a skeleton are employed to capture the scale-invariant spatial information of the skeleton. The static feature of the skeleton is defined as a combination of the orientations. Unlike previous orientation-based representations, the orientation of a rigid body in the skeleton is defined as the rotations between the rigid body and the coordinate axes in three-dimensional space. Each rotation is mapped to the special orthogonal group SO(3). Next, the rigid-body motions between the skeleton and its previous skeletons are utilized to capture the temporal information of the skeleton. The dynamic feature of the skeleton is defined as a combination of the motions. Similarly, the motions are represented as points in the special Euclidean group SE(3). Therefore, the proposed skeleton representation lies in the Lie group (SE(3)×⋯×SE(3), SO(3)×⋯×SO(3)), which is a manifold. Using the proposed representation, an action can be considered as a series of points in this Lie group. Then, to recognize human action more accurately, a new pattern-growth algorithm named MinP-PrefixSpan is proposed to mine the key-skeleton-patterns from the training dataset. Because the algorithm reduces the number of new patterns in each growth step, it is more efficient than the PrefixSpan algorithm. Finally, the key-skeleton-patterns are used to discover the most informative skeleton sequences of each action (skeleton sequence). Our approach achieves accuracies of 94.70%, 98.87%, and 95.01% on three action datasets, outperforming other relative action recognition approaches, including LieNet, Lie group, Grassmann manifold, and Graph-based model.


  • Robust visual tracking via randomly projected instance learning

    by Fei Cheng, Kai Liu, Mao-Guo Gong, Kaiyuan Fu, Jiangbo Xi

    Purpose: The purpose of this paper is to design a robust tracking algorithm which is suitable for the real-time requirement and solves the mistake labeling issue in the appearance model of trackers with the spare features. Design/methodology/approach: This paper proposes a tracker to select the most discriminative randomly projected ferns and integrates a coarse-to-fine search strategy in this framework. First, the authors exploit multiple instance boosting learning to maximize the bag likelihood and select randomly projected fern from feature pool to degrade the effect of mistake labeling. Second, a coarse-to-fine search approach is first integrated into the framework of multiple instance learning (MIL) for less detections. Findings: The quantitative and qualitative experiments demonstrate that the tracker has shown favorable performance in efficiency and effective among the competitors of tracking algorithms. Originality/value: The proposed method selects the feature from the compressive domain by MIL AnyBoost and integrates the coarse-to-fine search strategy first to reduce the burden of detection. This paper designs a tracker with high speed and favorable results which is more suitable for real-time scene.