Visual Recognition Problems Research Articles

Divide and Retain: A Dual-Phase Modeling for Long-Tailed Visual Recognition.

This work explores visual recognition models on real-world datasets exhibiting a long-tailed distribution. Most of previous works are based on a holistic perspective that the overall gradient for training model is directly obtained by considering all classes jointly. However, due to the extreme data imbalance in long-tailed datasets, joint consideration of different classes tends to induce the gradient distortion problem; i.e., the overall gradient tends to suffer from shifted direction toward data-rich classes and enlarged variances caused by data-poor classes. The gradient distortion problem impairs the training of our models. To avoid such drawbacks, we propose to disentangle the overall gradient and aim to consider the gradient on data-rich classes and that on data-poor classes separately. We tackle the long-tailed visual recognition problem via a dual-phase-based method. In the first phase, only data-rich classes are concerned to update model parameters, where only separated gradient on data-rich classes is used. In the second phase, the rest data-poor classes are involved to learn a complete classifier for all classes. More importantly, to ensure the smooth transition from phase I to phase II, we propose an exemplar bank and a memory-retentive loss. In general, the exemplar bank reserves a few representative examples from data-rich classes. It is used to maintain the information of data-rich classes when transiting. The memory-retentive loss constrains the change of model parameters from phase I to phase II based on the exemplar bank and data-poor classes. The extensive experimental results on four commonly used long-tailed benchmarks, including CIFAR100-LT, Places-LT, ImageNet-LT, and iNaturalist 2018, highlight the excellent performance of our proposed method.

Learning From Human Attention for Attribute-Assisted Visual Recognition.

With prior knowledge of seen objects, humans have a remarkable ability to recognize novel objects using shared and distinct local attributes. This is significant for the challenging tasks of zero-shot learning (ZSL) and fine-grained visual classification (FGVC), where the discriminative attributes of objects have played an important role. Inspired by human visual attention, neural networks have widely exploited the attention mechanism to learn the locally discriminative attributes for challenging tasks. Though greatly promoted the development of these fields, existing works mainly focus on learning the region embeddings of different attribute features and neglect the importance of discriminative attribute localization. It is also unclear whether the learned attention truly matches the real human attention. To tackle this problem, this paper proposes to employ real human gaze data for visual recognition networks to learn from human attention. Specifically, we design a unified Attribute Attention Network (A 2 Net) that learns from human attention for both ZSL and FGVC tasks. The overall model consists of an attribute attention branch and a baseline classification network. On top of the image feature maps provided by the baseline classification network, the attribute attention branch employs attribute prototypes to produce attribute attention maps and attribute features. The attribute attention maps are converted to gaze-like attentions to be aligned with real human gaze attention. To guarantee the effectiveness of attribute feature learning, we further align the extracted attribute features with attribute-defined class embeddings. To facilitate learning from human gaze attention for the visual recognition problems, we design a bird classification game to collect real human gaze data using the CUB dataset via an eye-tracker device. Experiments on ZSL and FGVC tasks without/with real human gaze data validate the benefits and accuracy of our proposed model. This work supports the promising benefits of collecting human gaze datasets and automatic gaze estimation algorithms learning from human attention for high-level computer vision tasks.

Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data

Learning from open-world noisy data, where both closed-set and open-set noise co-exist in the dataset, is a realistic but underexplored setting. Only recently, several efforts have been initialized to tackle this problem. However, these works assume the classes are balanced when dealing with open-world noisy data. This assumption often violates the nature of real-world large-scale datasets, where the label distributions are generally long-tailed, i.e. class-imbalanced. In this paper, we study the problem of robust visual recognition with class-imbalanced open-world noisy data. We propose a probabilistic graphical model-based approach: iMRF to achieve label noise correction that is robust to class imbalance via an efficient iterative inference of a Markov Random Field (MRF) in each training mini-batch. Furthermore, we design an agreement-based thresholding strategy to adaptively collect clean samples from all classes that includes corrected closed-set noisy samples while rejecting open-set noisy samples. We also introduce a noise-aware balanced cross-entropy loss to explicitly eliminate the bias caused by class-imbalanced data. Extensive experiments on several benchmark datasets including synthetic and real-world noisy datasets demonstrate the superior performance robustness of our method over existing methods. Our code is available at https://github.com/Na-Z/LIOND.

Parallel Self-Attention and Spatial-Attention Fusion for Human Pose Estimation and Running Movement Recognition

Human pose estimation (HPE) is a fundamental yet promising visual recognition problem. Existing popular methods (e.g., Hourglass and its variants) either attempt to directly add local features element-wisely, or (e.g., vision transformers) try to learn the global relationships among different human parts. However, it remains an open problem to effectively integrate the local-global representations for accurate HPE. In this work, we design four feature fusion strategies on the hierarchical ResNet structure, including direct channel concatenation, element-wise addition and two parallel structures. Both two parallel structures adopt the naive self-attention encoder to model global dependencies. The difference between them is that one adopts the original ResNet BottleNeck while the other employs a spatial-attention module (named SSF) to learn the local patterns. Experiments on COCO Keypoint 2017 show that our SSF for HPE (named SSPose) achieves the best average precision with acceptable computational cost among the compared state-of-the-art methods. In addition, we build a lightweight running dataset to verify the effectiveness of SSPose. Based solely on the keypoints estimated by our SSPose, we propose a regression model to identify valid running movements without training any other classifiers. Our source codes and running dataset are publicly available.

Attentional Composition Networks for Long-Tailed Human Action Recognition

The problem of long-tailed visual recognition has been receiving increasing research attention. However, the long-tailed distribution problem remains underexplored for video-based visual recognition. To address this issue, in this article we propose a compositional learning based solution for video-based human action recognition. Our method, named Attentional Composition Networks (ACN), first learns verb-like and preposition-like components, then shuffles these components to generate samples for the tail classes in the feature space to augment the data for the tail classes. Specifically, during training, we represent each action video by a graph that captures the spatial-temporal relations (edges) among detected human/object instances (nodes). Then, ACN utilizes the position information to decompose each action into a set of verb and preposition representations using the edge features in the graph. After that, the verb and preposition features from different videos are combined via an attention structure to synthesize feature representations for tail classes. This way, we can enrich the data for the tail classes and consequently improve the action recognition for these classes. To evaluate the compositional human action recognition, we further contribute a new human action recognition dataset, namely NEU-Interaction (NEU-I). Experimental results on both Something-Something V2 and the proposed NEU-I demonstrate the effectiveness of the proposed method for long-tailed, few-shot, and zero-shot problems in human action recognition. Source code and the NEU-I dataset are available at https://github.com/YajieW99/ACN .

Semantic-guided de-attention with sharpened triplet marginal loss for visual place recognition

Thanks to Earth-level Street View images from Google Maps, a visual image geo-localization can estimate the coarse location of a query image with a visual place recognition process. However, this can get very challenging when non-static objects change with time, severely degrading image retrieval accuracy. We address the problem of city-scale visual place recognition in complex urban environments crowded with non-static clutters. To this end, we first analyze what clutters degrade similarity matching between the query and database images. Second, we design a self-supervised trainable de-attention module that prevents the network from focusing on non-static objects in an input image. In addition, we propose a novel triplet marginal loss called sharpened triplet marginal loss to make feature descriptors more discriminative. Lastly, due to the lack of geo-tagged public datasets with a high density of non-static objects, we propose a clutter augmentation method to evaluate our approach. The experimental results show that our model has notably improved over the existing attention methods in geo-localization tasks on the public benchmark datasets and on their augmented versions with high population and traffic. Our code is available at https://github.com/ccsmm78/deattention_with_stml_for_vpr.

HMLNet: a hierarchical metric learning network with dual attention for change detection in high-resolution remote sensing images

ABSTRACT Change detection (CD) in high-resolution remote sensing images (RSIs) can be regarded as a binary visual recognition problem. Metric learning (ML) is a reliable method to determine pixel class attributes based on discriminative distance function between learnable image features. However, most related works learn discriminative distance functions in single-scale feature pairs, which suffer from slow convergence and poor local optima, partially due to that the loss function employs only large-scale feature samples while not interacting with the other scale features. Furthermore, more effective features are a prerequisite for improving the performance of ML-based CD. Hence, we propose a novel hierarchical metric learning network (HMLNet) with dual attention for CD in RSIs, where the key point is that hierarchical metric learning is performed in an ensemble manner to improve detection accuracy and accelerate model convergence. Specifically, based on the features extracted from the encoder-decoder backbone, we construct a feature pyramid to handle the complex details of objects at various scales in RSIs, and then perform metric learning between the paired pyramid features at the same scale. In addition, the dual-attention module is proposed to enhance the internal consistency of changed objects and effectively obtain more detailed information by acting on multi-scale pyramid features. Extensive experiments on the two public RSIs CD datasets, and the results demonstrate that the proposed HMLNet can accurately locate changed objects, which consistently outperforms the state-of-the-art CD competitors.

Unpaired Image Captioning by Image-Level Weakly-Supervised Visual Concept Recognition

The goal of unpaired image captioning (UIC) is to describe images without using image-caption pairs in the training phase. Although challenging, we except the task can be accomplished by leveraging a training set of images aligned with visual concepts. Most existing studies use off-the-shelf algorithms to obtain the visual concepts because the Bounding Box (BBox) labels or relationship-triplet labels used for the training are expensive to acquire. In order to resolve the problem in expensive annotations, we propose a novel approach to achieve cost-effective UIC. Specifically, we adopt image-level labels for the optimization of the UIC model in a weakly-supervised manner. For each image, we assume that only the image-level labels (such as object categories and the relationships) are available without specific locations and numbers. The image-level labels are utilized to train a weakly-supervised object recognition model to extract object information (e.g., instance) in an image, and the extracted instances are adopted to infer the relationships among different objects based on an enhanced graph neural network (GNN). The proposed approach achieves comparable or even better performance compared with previous methods without the expensive cost of annotations. Furthermore, we design an unrecognized object (UnO) loss combined with a visual concept reward to improve the alignment of the inferred object and relationship information with the images. It can effectively alleviate the issue encountered by existing UIC models about generating sentences with nonexistent objects. To the best of our knowledge, this is the first attempt to solve the problem of Weakly-Supervised visual concept recognition for UIC (WS-UIC) based only on image-level labels. Extensive experiments have been carried out to demonstrate that the proposed WS-UIC model achieves inspiring results on the COCO dataset while significantly reducing the cost of labeling.

Toward Deep Transfer Learning for Realistic Activity Recognition in Videos

Today, videos have become popular on the internet and specified in social network and media platforms such as Youbue, Ticktok, and Vimeo. Video understanding has attracted much attention in the research community in recent years. Automatically recognizing human activity in wild videos is a trending research topic with a wide range of applications in advertising, smarthome, and surveillance camera systems. Deep convolutional neural networks have become a new de facto visual recognition problem. It achieved much success in image recognition problems that are leveraged from the ImageNet dataset. Many researchers have applied CNNs to the video domain, but the results in realistic video still have many challenges, and the recognition rate is not as expected. Because the realistic activity in video is extremely small, we cannot train a large deep convolutional network to achieve good performance. In this work, we answer the question “how could we transfer the visual features from the ImageNet dataset into video for activity recognition tasks?”. We propose an approach based on the pretrained models from ImageNet for activity recognition in video that is based on transforming video into an image map for temporal information and a frame-based description for spatial information. We test our proposed approach on three datasets, UCF11, UCF50 and UCF101, and it achieves an accuracy of over 95% for the backbone of ResNet and over 88% for the backbone of MobileNet. The experimental results show that our proposed method is robust and efficient in wild video datasets.

Research on the Application of Visual Recognition in the Engine Room of Intelligent Ships.

In the engine room of intelligent ships, visual recognition is an essential technical precondition for automatic inspection. At present, the problems of visual recognition in marine engine rooms include missing detection, low accuracy, slow speed, and imperfect datasets. For these problems, this paper proposes a marine engine room equipment recognition model based on the improved You Only Look Once v5 (YOLOv5) algorithm. The channel pruning method based on batch normalization (BN) layer weight value is used to improve the recognition speed. The complete intersection over union (CIoU) loss function and hard-swish activation function are used to enhance detection accuracy. Meanwhile, soft-NMS is used as the non-maximum suppression (NMS) method to reduce the false rate and missed detection rate. Then, the main equipment in the marine engine room (MEMER) dataset is built. Finally, comparative experiments and ablation experiments are carried out on the MEMER dataset to verify the strategy’s efficacy on the model performance boost. Specifically, this model can accurately detect 100.00% of diesel engines, 95.91% of pumps, 94.29% of coolers, 98.54% of oil separators, 64.21% of meters, 60.23% of reservoirs, and 75.32% of valves in the actual marine engine room.

Binary Neural Networks for Memory-Efficient and Effective Visual Place Recognition in Changing Environments

Visual place recognition (VPR) is a robot's ability to determine whether a place was visited before using visual data. While conventional hand-crafted methods for VPR fail under extreme environmental appearance changes, those based on convolutional neural networks (CNNs) achieve state-of-the-art performance but result in heavy runtime processes and model sizes that demand a large amount of memory. Hence, CNN-based approaches are unsuitable for resource-constrained platforms, such as small robots and drones. In this paper, we take a multi-step approach of decreasing the precision of model parameters, combining it with network depth reduction and fewer neurons in the classifier stage to propose a new class of highly compact models that drastically reduces the memory requirements and computational effort while maintaining state-of-the-art VPR performance. To the best of our knowledge, this is the first attempt to propose binary neural networks for solving the visual place recognition problem effectively under changing conditions and with significantly reduced resource requirements. Our best-performing binary neural network, dubbed FloppyNet, achieves comparable VPR performance when considered against its full-precision and deeper counterparts while consuming 99% less memory and increasing the inference speed seven times.

Multi-View Graph Fusion for Semi-Supervised Learning: Application to Image-Based Face Beauty Prediction

Facial Beauty Prediction (FBP) is an important visual recognition problem to evaluate the attractiveness of faces according to human perception. Most existing FBP methods are based on supervised solutions using geometric or deep features. Semi-supervised learning for FBP is an almost unexplored research area. In this work, we propose a graph-based semi-supervised method in which multiple graphs are constructed to find the appropriate graph representation of the face images (with and without scores). The proposed method combines both geometric and deep feature-based graphs to produce a high-level representation of face images instead of using a single face descriptor and also improves the discriminative ability of graph-based score propagation methods. In addition to the data graph, our proposed approach fuses an additional graph adaptively built on the predicted beauty values. Experimental results on the SCUTFBP-5500 facial beauty dataset demonstrate the superiority of the proposed algorithm compared to other state-of-the-art methods.

Prototype-based classifier learning for long-tailed visual recognition

In this paper, we tackle the long-tailed visual recognition problem from the categorical prototype perspective by proposing a prototype-based classifier learning (PCL) method. Specifically, thanks to the generalization ability and robustness, categorical prototypes reveal their advantages of representing the category semantics. Coupled with their class-balance characteristic, categorical prototypes also show potential for handling data imbalance. In our PCL, we propose to generate the categorical classifiers based on the prototypes by performing a learnable mapping function. To further alleviate the impact of imbalance on classifier generation, two kinds of classifier calibration approaches are designed from both prototype-level and example-level aspects. Extensive experiments on five benchmark datasets, including the large-scale iNaturalist, Places-LT, and ImageNet-LT, justify that the proposed PCL can outperform state-of-the-arts. Furthermore, validation experiments can demonstrate the effectiveness of tailored designs in PCL for long-tailed problems.

Brain-like functional specialization emerges spontaneously in deep neural networks.

The human brain contains multiple regions with distinct, often highly specialized functions, from recognizing faces to understanding language to thinking about what others are thinking. However, it remains unclear why the cortex exhibits this high degree of functional specialization in the first place. Here, we consider the case of face perception using artificial neural networks to test the hypothesis that functional segregation of face recognition in the brain reflects a computational optimization for the broader problem of visual recognition of faces and other visual categories. We find that networks trained on object recognition perform poorly on face recognition and vice versa and that networks optimized for both tasks spontaneously segregate themselves into separate systems for faces and objects. We then show functional segregation to varying degrees for other visual categories, revealing a widespread tendency for optimization (without built-in task-specific inductive biases) to lead to functional specialization in machines and, we conjecture, also brains.

DeepStreamOS: Fast open-Set classification for convolutional neural networks

Convolutional Neural Networks (CNNs) can achieve state of the art results for visual recognition problems when the train and test data distributions are the same and when all classes in the test set are present in the training data. This is not representative of the real-world where the data evolves; existing classes change and new classes emerge. A traditional neural network only has the capacity to label instances with classes it has been trained on and cannot identify unknown classes. This can be of serious consequence in safety critical systems. The research field of Open-Set Classification provides potential solutions to overcome the identification of unknown classes in deep neural networks. In safety-critical systems, the speed with which unknown classes can be identified is also essential. Our system, termed DeepStreamOS, brings together the use of deep neural network activations with a stream-based outlier detection method for fast identification of instances that belong to unknown classes. DeepStreamOS uses all layers of a CNN to get a trajectory of the activations and applies a stream-based analysis method to determine if an instance belongs to an unknown class. We use CIFAR-10 and Fashion-MNIST datasets, withholding classes to apply as unknown data on VGG16 and MobileNet deep neural networks. Our system is compared with leading open-set classification methods: OpenMax and EVM. We show that DeepStreamOS outperforms OpenMax and EVM in most open-set classification scenarios and by a large margin on speed in all scenarios.

GSAP: A Global Structure Attention Pooling Method for Graph-Based Visual Place Recognition

The Visual Place Recognition problem aims to use an image to recognize the location that has been visited before. In most of the scenes revisited, the appearance and view are drastically different. Most previous works focus on the 2-D image-based deep learning method. However, the convolutional features are not robust enough to the challenging scenes mentioned above. In this paper, in order to take advantage of the information that helps the Visual Place Recognition task in these challenging scenes, we propose a new graph construction approach to extract the useful information from an RGB image and a depth image and fuse them in graph data. Then, we deal with the Visual Place Recognition problem as a graph classification problem. We propose a new Global Pooling method—Global Structure Attention Pooling (GSAP), which improves the classification accuracy by improving the expression ability of the Global Pooling component. The experiments show that our GSAP method improves the accuracy of graph classification by approximately 2–5%, the graph construction method improves the accuracy of graph classification by approximately 4–6%, and that the whole Visual Place Recognition model is robust to appearance change and view change.

Towards a Robust Visual Place Recognition in Large-Scale vSLAM Scenarios Based on a Deep Distance Learning.

The application of deep learning is blooming in the field of visual place recognition, which plays a critical role in visual Simultaneous Localization and Mapping (vSLAM) applications. The use of convolutional neural networks (CNNs) achieve better performance than handcrafted feature descriptors. However, visual place recognition is still a challenging task due to two major problems, i.e., perceptual aliasing and perceptual variability. Therefore, designing a customized distance learning method to express the intrinsic distance constraints in the large-scale vSLAM scenarios is of great importance. Traditional deep distance learning methods usually use the triplet loss which requires the mining of anchor images. This may, however, result in very tedious inefficient training and anomalous distance relationships. In this paper, a novel deep distance learning framework for visual place recognition is proposed. Through in-depth analysis of the multiple constraints of the distance relationship in the visual place recognition problem, the multi-constraint loss function is proposed to optimize the distance constraint relationships in the Euclidean space. The new framework can support any kind of CNN such as AlexNet, VGGNet and other user-defined networks to extract more distinguishing features. We have compared the results with the traditional deep distance learning method, and the results show that the proposed method can improve the performance by 19–28%. Additionally, compared to some contemporary visual place recognition techniques, the proposed method can improve the performance by 40%/36% and 27%/24% in average on VGGNet/AlexNet using the New College and the TUM datasets, respectively. It’s verified the method is capable to handle appearance changes in complex environments.

Plant Disease Recognition: A Large-Scale Benchmark Dataset and a Visual Region and Loss Reweighting Approach.

Plant disease diagnosis is very critical for agriculture due to its importance for increasing crop production. Recent advances in image processing offer us a new way to solve this issue via visual plant disease analysis. However, there are few works in this area, not to mention systematic researches. In this paper, we systematically investigate the problem of visual plant disease recognition for plant disease diagnosis. Compared with other types of images, plant disease images generally exhibit randomly distributed lesions, diverse symptoms and complex backgrounds, and thus are hard to capture discriminative information. To facilitate the plant disease recognition research, we construct a new large-scale plant disease dataset with 271 plant disease categories and 220,592 images. Based on this dataset, we tackle plant disease recognition via reweighting both visual regions and loss to emphasize diseased parts. We first compute the weights of all the divided patches from each image based on the cluster distribution of these patches to indicate the discriminative level of each patch. Then we allocate the weight to each loss for each patch-label pair during weakly-supervised training to enable discriminative disease part learning. We finally extract patch features from the network trained with loss reweighting, and utilize the LSTM network to encode the weighed patch feature sequence into a comprehensive feature representation. Extensive evaluations on this dataset and another public dataset demonstrate the advantage of the proposed method. We expect this research will further the agenda of plant disease recognition in the community of image processing.

Survey on the Analysis and Modeling of Visual Kinship: A Decade in the Making.

Kinship recognition is a challenging problem with many practical applications. With much progress and milestones having been reached after ten years, we are now able to survey the research and create new milestones. We review the public resources and data challenges that enabled and inspired many to hone-in on the views of automatic kinship recognition in the visual domain. The different tasks are described in technical terms and syntax consistent across the problem domain and the practical value of each discussed and measured. State-of-the-art methods for visual kinship recognition problems, whether to discriminate between or generate from, are examined. As part of such, we review systems proposed as part of a recent data challenge held in conjunction with the 2020 IEEE Conference on Automatic Face and Gesture Recognition. We establish a stronghold for the state of progress for the different problems in a consistent manner. This survey will serve as the central resource for the work of the next decade to build upon. For the tenth anniversary, the demo code is provided for the various kin-based tasks. Detecting relatives with visual recognition and classifying the relationship is an area with high potential for impact in research and practice.

Transfer learning with deep convolutional neural network for automated plant identification in hainan island

The Hainan Island has a generally high biological diversity with a wide variety of plant species, some of which are listed as endemic to the island. It is time-consuming and difficult, even for the botanist experts to determine the name of species based on observations. Automated plant identification enables experts to process significantly greater numbers of plants with higher efficiencies in shorter periods of time. However, plant recognition is a kind of fine-grained visual recognition problem, which is relatively harder than conventional image recognition. In this paper, we employ a Deep Convolutional Neural Network (DCNN) trained on the ImageNet database, which contains millions of images, and then transfer the learning information for automated plant identification based on flower and fruit images. First, we modify the last three layers of the pre-trained network in order to adapt ResNet-50 model to our classification task, and replace the fully connected layer in the original pre-trained network with another fully connected layers, in which the output size represents the class of plants. Secondly, we use transfer experience and fine-tuned pre-trained DCNN for experiments using flower and fruit images. Finally, we evaluate the proposed network on two available botanical datasets: the Oxford flowers dataset with 102 classes and the HNPlant flowers and fruits dataset with 20 classes, and determine the optimal values of the associated hyperparameters to improve the overall performance. Experiment results demonstrate that the highest classification accuracies exhibited by the proposed model on the Oxford-102 and HNPlant-20 datasets are 92.4% and 95.0%, respectively, thus establishing their effectiveness and superiority.