Due to the extreme imbalance in the number of normal data and abnormal data, visual anomaly detection is important for the development of industrial automatic product quality inspection. Unsupervised methods based on reconstruction and embedding have been widely studied for anomaly detection, of which reconstruction-based methods are the most popular. However, establishing a unified model for textured surface defect detection remains a challenge because these surfaces can vary in homogeneous and non regularly ways. Furthermore, existing reconstruction-based methods do not have a strong ability to convert the defect feature to the normal feature. To address these challenges, we propose a novel unsupervised reference-based autoencoder (RB-AE) to accurately inspect a variety of textured defects. Unlike most reconstruction-based methods, artificial defects and a novel pixel-level discrimination loss function are utilized for training to enable the model to obtain pixel-level discrimination ability. First, the RB-AE employs an encoding module to extract multi-scale features of the textured surface. Subsequently, a novel reference-based attention module (RBAM) is proposed to convert the defect features to normal features to suppress the reconstruction of defects. In addition, RBAM can also effectively suppress the defective feature residual caused by skip-connection. Next, a decoding module utilizes the repaired features to reconstruct the normal texture background. Finally, a novel multiscale feature discrimination module (MSFDM) is employed to defect detection and segmentation.
End-to-end Speech Translation (ST) aims at translating the source language speech into target language text without generating the intermediate transcriptions. However, the training of end-to-end methods relies on parallel ST data, which are difficult and expensive to obtain. Fortunately, the supervised data for automatic speech recognition (ASR) and machine translation (MT) are usually more accessible, making zero-shot speech translation a potential direction. Existing zero-shot methods fail to align the two modalities of speech and text into a shared semantic space, resulting in much worse performance compared to the supervised ST methods. In order to enable zero-shot ST, we propose a novel Discrete Cross-Modal Alignment (DCMA) method that employs a shared discrete vocabulary space to accommodate and match both modalities of speech and text. Specifically, we introduce a vector quantization module to discretize the continuous representations of speech and text into a finite set of virtual tokens, and use ASR data to map corresponding speech and text to the same virtual token in a shared codebook. This way, source language speech can be embedded in the same semantic space as the source language text, which can be then transformed into target language text with an MT module. Experiments on multiple language pairs demonstrate that our zero-shot ST method significantly improves the SOTA, and even performers on par with the strong supervised ST baselines.
Planning a time-optimal trajectory for aerial robots is critical in many drone applications, such as rescue missions and package delivery, which have been widely researched in recent years. However, it still involves several challenges, particularly when it comes to incorporating special task requirements into the planning as well as the aerial robot's dynamics. In this work, we study a case where an aerial manipulator shall hand over a parcel from a moving mobile robot in a time-optimal manner. Rather than setting up the approach trajectory manually, which makes it difficult to determine the optimal total travel time to accomplish the desired task within dynamic limits, we propose an optimization framework, which combines discrete mechanics and complementarity constraints (DMCC) together. In the proposed framework, the system dynamics is constrained with the discrete variational Lagrangian mechanics that provides reliable estimation results also according to our experiments. The handover opportunities are automatically determined and arranged based on the desired complementarity constraints. Finally, the performance of the proposed framework is verified with numerical simulations and hardware experiments with our self-designed aerial manipulators.
In surface defect detection, due to the extreme imbalance in the number of positive and negative samples, positive-samples-based anomaly detection methods have received more and more attention. Specifically, reconstruction-based methods are the most popular. However, exiting methods are either difficult to repair abnormal foregrounds or reconstruct clear backgrounds. Therefore, we propose a clear memory-augmented auto-encoder. At first, we propose a novel clear memory-augmented module, which combines the encoding and memory-encoding in a way of forgetting and inputting, thereby repairing abnormal foregrounds and preservation clear backgrounds. Secondly, a general artificial anomaly generation algorithm is proposed to simulate anomalies that are as realistic and feature-rich as possible. At last, we propose a novel multi scale feature residual detection method for defect segmentation, which makes the defect location more accurate. CMA-AE conducts comparative experiments using 11 state-of-the-art methods on five benchmark datasets, showing an average 18.6% average improvement in F1-measure.
Effectively classifying remote sensing scenes is still a challenge due to the increasing spatial resolution of remote imaging and large variances between remote sensing images. Existing research has greatly improved the performance of remote sensing scene classification (RSSC). However, these methods are not applicable to cross-domain few-shot problems where target domain is with very limited training samples available and has a different data distribution from source domain. To improve the model's applicability, we propose the feature-wise transformation module (FTM) in this paper. FTM transfers the feature distribution learned on source domain to that of target domain by a very simple affine operation with negligible additional parameters. Moreover, FTM can be effectively learned on target domain in the case of few training data available and is agnostic to specific network structures. Experiments on RSSC and land-cover mapping tasks verified its capability to handle cross-domain few-shot problems. By comparison with directly finetuning, FTM achieves better performance and possesses better transferability and fine-grained discriminability. \textit{Code will be publicly available.}
Deep neural networks tend to underestimate uncertainty and produce overly confident predictions. Recently proposed solutions, such as MC Dropout and SDENet, require complex training and/or auxiliary out-of-distribution data. We propose a simple solution by extending the time-tested iterative reweighted least square (IRLS) in generalised linear regression. We use two sub-networks to parametrise the prediction and uncertainty estimation, enabling easy handling of complex inputs and nonlinear response. The two sub-networks have shared representations and are trained via two complementary loss functions for the prediction and the uncertainty estimates, with interleaving steps as in a cooperative game. Compared with more complex models such as MC-Dropout or SDE-Net, our proposed network is simpler to implement and more robust (insensitive to varying aleatoric and epistemic uncertainty).
In recent years, Graph Convolutional Networks (GCNs) have achieved great success in learning from graph-structured data. With the growing tendency of graph nodes and edges, GCN training by single processor cannot meet the demand for time and memory, which led to a boom into distributed GCN training frameworks research. However, existing distributed GCN training frameworks require enormous communication costs between processors since multitudes of dependent nodes and edges information need to be collected and transmitted for GCN training from other processors. To address this issue, we propose a Graph Augmentation based Distributed GCN framework(GAD). In particular, GAD has two main components, GAD-Partition and GAD-Optimizer. We first propose a graph augmentation-based partition (GAD-Partition) that can divide original graph into augmented subgraphs to reduce communication by selecting and storing as few significant nodes of other processors as possible while guaranteeing the accuracy of the training. In addition, we further design a subgraph variance-based importance calculation formula and propose a novel weighted global consensus method, collectively referred to as GAD-Optimizer. This optimizer adaptively reduces the importance of subgraphs with large variances for the purpose of reducing the effect of extra variance introduced by GAD-Partition on distributed GCN training. Extensive experiments on four large-scale real-world datasets demonstrate that our framework significantly reduces the communication overhead (50%), improves the convergence speed (2X) of distributed GCN training, and slight gain in accuracy (0.45%) based on minimal redundancy compared to the state-of-the-art methods.
Translating e-commercial product descriptions, a.k.a product-oriented machine translation (PMT), is essential to serve e-shoppers all over the world. However, due to the domain specialty, the PMT task is more challenging than traditional machine translation problems. Firstly, there are many specialized jargons in the product description, which are ambiguous to translate without the product image. Secondly, product descriptions are related to the image in more complicated ways than standard image descriptions, involving various visual aspects such as objects, shapes, colors or even subjective styles. Moreover, existing PMT datasets are small in scale to support the research. In this paper, we first construct a large-scale bilingual product description dataset called Fashion-MMT, which contains over 114k noisy and 40k manually cleaned description translations with multiple product images. To effectively learn semantic alignments among product images and bilingual texts in translation, we design a unified product-oriented cross-modal cross-lingual model (\upoc~) for pre-training and fine-tuning. Experiments on the Fashion-MMT and Multi30k datasets show that our model significantly outperforms the state-of-the-art models even pre-trained on the same dataset. It is also shown to benefit more from large-scale noisy data to improve the translation quality. We will release the dataset and codes at https://github.com/syuqings/Fashion-MMT.
Link prediction plays an important role in network analysis and applications. Recently, approaches for link prediction have evolved from traditional similarity-based algorithms into embedding-based algorithms. However, most existing approaches fail to exploit the fact that real-world networks are different from random networks. In particular, real-world networks are known to contain motifs, natural network building blocks reflecting the underlying network-generating processes. In this paper, we propose a novel embedding algorithm that incorporates network motifs to capture higher-order structures in the network. To evaluate its effectiveness for link prediction, experiments were conducted on three types of networks: social networks, biological networks, and academic networks. The results demonstrate that our algorithm outperforms both the traditional similarity-based algorithms by 20% and the state-of-the-art embedding-based algorithms by 19%.
We aim to provide a computationally cheap yet effective approach for fine-grained image classification (FGIC) in this letter. Unlike previous methods that rely on complex part localization modules, our approach learns fine-grained features by improving the semantics of sub-features of a global feature. Specifically, we first achieve the sub-feature semantic by arranging feature channels of a CNN into different groups through channel permutation. Meanwhile, to enhance the discriminability of sub-features, the groups are guided to be activated on object parts with strong discriminability by a weighted combination regularization. This process brings only 1.7% additional parameters to its ResNet-50 backbone. Moreover, our approach can be easily integrated into the backbone model as a plug-and-play module for end-to-end training with only image-level supervision. Experiments verified the effectiveness of our approach and validated its comparable performance to the state-of-the-art methods. Code is available at https://github.com/cswluo/SEF