Object Class Recognition Research Articles

Enhancing functional efficiency in information-extreme machine learning with logistic regression ensembles

The subject matter of the article is the application of supervised machine learning for the task of object class recognition. The goal is enhancing functional efficiency in information-extreme technology (IET) for object class recognition. The tasks to be solved are: to analyse possible ways of increasing the functional efficiency IET approach; implement an ensemble of models that include logistic regression for prioritizing recognition features and an IEI learning algorithm; compare the functional efficiency of the resulting ensemble of models on well-known dataset with classic approach and results of other researchers. The methods: The method is developed within the framework of the functional approach to modelling natural intelligence applied to the problem of object classification. The following results were obtained: The study tries to augment existing IET to support feature prioritization as part of the object class recognition algorithm. The classical information-extreme algorithm treats all input features equivalently important in forming the decisive rule. As a result, the object features with strong correlation are not prioritized by the algorithm's decisive mechanism – resulting in decreasing functional efficiency in exam mode. The proposed approach is solving this problem by applying a two-stage approach. In the first stage the multiclass logistic regression applied to the input training features vectors of objects to be classified – formed the normalized training matrix. To prevent overfitting of the logistic regression, a model the L2(ridge) regularization method was used. On the second stage, the information-extreme method as input takes the result of the first stage. The geometrical parameters of class containers and the control tolerances on the recognition features were considered as the optimization parameters. Conclusions. The proposed approach increases MNIST (Modified National Institute of Standards and Technology) dataset classification accuracy compared with the classic information-extreme method by 26,44%. The proposed approach has a 3.77% lower accuracy compared to neural-like approaches but uses fewer resources in the training phase and allows retraining the model, as well as expanding the dictionary of recognition classes without model retraining.

Exploring pattern recognition: what is the relationship between the recognition of words, faces and other objects?

Debate surrounds processes of visual recognition, with no consensus as to whether recognition of distinct object categories (faces, bodies, cars, and words) is domain specific or subserved by domain-general visual recognition mechanisms. Here, we investigated correlations between the performance of 74 participants on recognition tasks for words, faces and other object categories. Participants completed a counter-balanced test battery of the Cambridge Face, Car and Body Parts Memory tests, as well as a standard four category lexical decision task, with response time and recognition accuracy as dependent variables. Results revealed significant correlations across domains for both recognition accuracy and response time, providing some support for domain-general pattern recognition. Further exploration of the data using principal component analysis (PCA) revealed a two-component model for both the response time and accuracy data. However, how the various word and object recognition tasks fitted these components varied considerably but did hint at familiarity/expertise as a common factor. In sum, we argue a complex relationship exists between domain-specific processing and domain-general processing, but that this is shaped by expertise. To further our understanding of pattern recognition, research investigating the recognition of words, faces and other objects in dyslexic individuals is recommended, as is research exploiting neuroimaging methodologies, with excellent temporal resolution, to chart the temporal specifics of different forms of visual pattern recognition.

Remarkable archerfish recognise that unfamiliar insects are animals

Lurking beneath overhanging foliage, archerfish have one thing on their mind: taking a well-aimed pot-shot at the next insect that settles within range. Squirting a ballistic jet of water, the tenacious assassins precisely target their victims, ready to dine. ‘Object recognition is critical for animal survival’, says Ronen Segev from Ben-Gurion University of the Negev, Israel, explaining that distinguishing dinner is an essential life skill. But how do these remarkable fish differentiate a tasty treat from an unpalatable plant? ‘It's a complex task’, says Segev, who was curious to find out which visual features archerfish depend on when identifying a delicious nibble. He and his colleagues, Svetlana Volotsky, Ohad Ben-Shahar and Opher Donchin, also from Ben-Gurion University of the Negev, began showing images of appetising insects, flowers and leaves to archerfish to find out how the aquatic snipers distinguish fauna from flora.Fortunately, the fish are quite content to squirt at images on a computer screen, so Segev and Volotsky first encouraged the archerfish to take fire at a circle on a screen, in return for a food pellet reward. Then, Volotsky trained the snipers to aim at images of one specific spider viewed from a range of angles while trying to distract it with an image of a piece of vegetation. Once the fish were reliably spitting at the spider images, she tested whether they were capable of recognising the spider from any random direction by offering the fish a completely new view of the spider, which they hadn't seen before, beside an image of a piece of vegetation or another spider. If the fish had really learned to recognise their spider, it would still take a shot at it, even if it was seeing it from an unfamiliar angle. Impressively, the archerfish were still able to select their target spider, regardless of the direction of view; they can definitely recognise individual objects. But could the fish learn from experience to categorise an insect that they had never seen before as an animal?This time, Volotsky trained the fish to spit at images of insects, ranging from ants and beetles to flies and spiders, while trying to distract them with images of plants. Once the fish were dependably taking shots at the insects, she tried outwitting them by presenting them with images of unfamiliar insects, to find out whether the fish could recognise that any old insect is an animal. Astoundingly, the fish still targeted the insect images, even though they had never seen those creatures before. ‘Archerfish can generalise from examples to make object recognition of natural object classes’, says Segev.But how were the determined fish able to learn to distinguish plants from animals? This time, Volotsky, Ben-Shahar, Donchin and Segev teamed up to break down images of insects, flowers and leaves into 18 different features. Then, they built a computer program (known as a support-vector machine), which can learn to classify different types of information, that reproduced the fish's decision-making process to find out which features are necessary for archerfish to distinguish between animals and plants. Impressively, the fish only needed six essential features, including the loose perimeter encircling the object, how jagged or smooth the shape is and the texture of its surface, to distinguish animals from plants, with the first two factors being the most significant. And this is quite remarkable because, apparently, we use a similar, albeit slightly more elaborate, strategy to recognise each other's faces.

Method of object class recognition based on the analysis of reflective characteristics

Method of object class recognition based on the analysis of reflective characteristics

Urdnet: A Cryo-EM Particle Automatic Picking Method

Cryo-Electron Microscopy (Cryo-EM) images are characterized by the low signal-to-noise ratio, low contrast, serious background noise, more impurities, less data, difficult data labeling, simpler image semantics, and relatively fixed structure, while U-Net obtains low resolution when downsampling rate information to complete object category recognition, obtains high-resolution information during upsampling to complete precise segmentation and positioning, fills in the underlying information through skip connection to improve the accuracy of image segmentation, and has advantages in biological image processing like Cryo-EM image. This article proposes A U-Net based residual intensive neural network (Urdnet), which combines point-level and pixel-level tags, used to accurately and automatically locate particles from cryo-electron microscopy images, and solve the bottleneck that cryo-EM Single-particle biological macromolecule reconstruction requires tens of thousands of automatically picked particles. The 80S ribosome, HCN1 channel and TcdA1 toxin subunits, and other public protein datasets have been trained and tested on Urdnet. The experimental results show that Urdnet could reach the same excellent particle picking performances as the mainstream methods of RELION, DeepPicker, and acquire the 3D structure of picked particles with higher resolution.

Two-Branch Attention Learning for Fine-Grained Class Incremental Learning

As a long-standing research area, class incremental learning (CIL) aims to effectively learn a unified classifier along with the growth of the number of classes. Due to the small inter-class variances and large intra-class variances, fine-grained visual categorization (FGVC) as a challenging visual task has not attracted enough attention in CIL. Therefore, the localization of critical regions specialized for fine-grained object recognition plays a crucial role in FGVC. Additionally, it is important to learn fine-grained features from critical regions in fine-grained CIL for the recognition of new object classes. This paper designs a network architecture named two-branch attention learning network (TBAL-Net) for fine-grained CIL. TBAL-Net can localize critical regions and learn fine-grained feature representation by a lightweight attention module. An effective training framework is proposed for fine-grained CIL by integrating TBAL-Net into an effective CIL process. This framework is tested on three popular fine-grained object datasets, including CUB-200-2011, FGVC-Aircraft, and Stanford-Car. The comparative experimental results demonstrate that the proposed framework can achieve the state-of-the-art performance on the three fine-grained object datasets.

Local-HDP: Interactive open-ended 3D object category recognition in real-time robotic scenarios

We introduce a non-parametric hierarchical Bayesian approach for open-ended 3D object categorization, named the Local Hierarchical Dirichlet Process (Local-HDP). This method allows an agent to learn independent topics for each category incrementally and to adapt to the environment in time. Each topic is a distribution of the visual words over a predefined dictionary. Using an inference algorithm, these latent variables are inferred from the dataset. Subsequently, the category of an object is determined based on the likelihood of generating a 3D object from the model. Hierarchical Bayesian approaches like Latent Dirichlet Allocation (LDA) can transform low-level features to high-level conceptual topics for 3D object categorization. However, the efficiency and accuracy of LDA-based approaches depend on the number of topics that is chosen manually. Moreover, fixing the number of topics for all categories can lead to overfitting or underfitting of the model. In contrast, the proposed Local-HDP can autonomously determine the number of topics for each category. Furthermore, the online variational inference method has been adapted for fast posterior approximation in the Local-HDP model. Experiments show that the proposed Local-HDP method outperforms other state-of-the-art approaches in terms of accuracy, scalability, and memory efficiency by a large margin. Moreover, two robotic experiments have been conducted to show the applicability of the proposed approach in real-time applications.

HCE: Hierarchical Context Embedding for Region-Based Object Detection.

State-of-the-art two-stage object detectors apply a classifier to a sparse set of object proposals, relying on region-wise features extracted by RoIPool or RoIAlign as inputs. The region-wise features, in spite of aligning well with the proposal locations, may still lack the crucial context information which is necessary for filtering out noisy background detections, as well as recognizing objects possessing no distinctive appearances. To address this issue, we present a simple but effective Hierarchical Context Embedding (HCE) framework, which can be applied as a plug-and-play component, to facilitate the classification ability of a series of region-based detectors by mining contextual cues. Specifically, to advance the recognition of context-dependent object categories, we propose an image-level categorical embedding module which leverages the holistic image-level context to learn object-level concepts. Then, novel RoI features are generated by exploiting hierarchically embedded context information beneath both whole images and interested regions, which are also complementary to conventional RoI features. Moreover, to make full use of our hierarchical contextual RoI features, we propose the early-and-late fusion strategies (i.e., feature fusion and confidence fusion), which can be combined to boost the classification accuracy of region-based detectors. Comprehensive experiments demonstrate that our HCE framework is flexible and generalizable, leading to significant and consistent improvements upon various region-based detectors, including FPN, Cascade R-CNN, Mask R-CNN and PA-FPN. With simple modification, our HCE framework can be conveniently adapted to fit the structure of one-stage detectors, and achieve improved performance for SSD, RetinaNet and EfficientDet.

Dynamic dot displays reveal material motion network in the human brain

There is growing research interest in the neural mechanisms underlying the recognition of material categories and properties. This research field, however, is relatively more recent and limited compared to investigations of the neural mechanisms underlying object and scene category recognition. Motion is particularly important for the perception of non-rigid materials, but the neural basis of non-rigid material motion remains unexplored. Using fMRI, we investigated which brain regions respond preferentially to material motion versus other types of motion. We introduce a new database of stimuli – dynamic dot materials – that are animations of moving dots that induce vivid percepts of various materials in motion, e.g. flapping cloth, liquid waves, wobbling jelly. Control stimuli were scrambled versions of these same animations and rigid three-dimensional rotating dots. Results showed that isolating material motion properties with dynamic dots (in contrast with other kinds of motion) activates a network of cortical regions in both ventral and dorsal visual pathways, including areas normally associated with the processing of surface properties and shape, and extending to somatosensory and premotor cortices. We suggest that such a widespread preference for material motion is due to strong associations between stimulus properties. For example viewing dots moving in a specific pattern not only elicits percepts of material motion; one perceives a flexible, non-rigid shape, identifies the object as a cloth flapping in the wind, infers the object's weight under gravity, and anticipates how it would feel to reach out and touch the material. These results are a first important step in mapping out the cortical architecture and dynamics in material-related motion processing.

Research progress of zero-shot learning

Although there have been encouraging breakthroughs in supervised learning since the renaissance of deep learning, the recognition of large-scale object classes remains a challenge, especially when some classes have no or few training samples. In this paper, the development of ZSL is reviewed comprehensively, including the evolution, key technologies, mainstream models, current research hotspots and future research directions. First, the evolution process is introduced from the perspectives of multi-shot, few-shot to zero-shot learning. Second, the key techniques of ZSL are analyzed in detail in terms of three aspects: visual feature extraction, semantic representation and visual-semantic mapping. Third, some typical models are interpreted in chronological order. Finally, closely related articles from the last three years are collected to analyze the current research hotspots and list future research directions.

Depth in convolutional neural networks solves scene segmentation.

Feed-forward deep convolutional neural networks (DCNNs) are, under specific conditions, matching and even surpassing human performance in object recognition in natural scenes. This performance suggests that the analysis of a loose collection of image features could support the recognition of natural object categories, without dedicated systems to solve specific visual subtasks. Research in humans however suggests that while feedforward activity may suffice for sparse scenes with isolated objects, additional visual operations ('routines') that aid the recognition process (e.g. segmentation or grouping) are needed for more complex scenes. Linking human visual processing to performance of DCNNs with increasing depth, we here explored if, how, and when object information is differentiated from the backgrounds they appear on. To this end, we controlled the information in both objects and backgrounds, as well as the relationship between them by adding noise, manipulating background congruence and systematically occluding parts of the image. Results indicate that with an increase in network depth, there is an increase in the distinction between object- and background information. For more shallow networks, results indicated a benefit of training on segmented objects. Overall, these results indicate that, de facto, scene segmentation can be performed by a network of sufficient depth. We conclude that the human brain could perform scene segmentation in the context of object identification without an explicit mechanism, by selecting or "binding" features that belong to the object and ignoring other features, in a manner similar to a very deep convolutional neural network.

Development of an "object category recognition" task for mice: Involvement of muscarinic acetylcholine receptors.

Recent research suggests that rats are capable of object categorization-like processes. To study whether mice possess similar abilities, we developed a spontaneous one-trial object category recognition (OCR) task. Based on the spontaneous object recognition paradigm, mice discriminated between two otherwise equally novel objects, one from a novel category and one from a studied category. During the sample phase, mice were exposed to two different exemplars from the same category. After a retention delay, they explored a third (i.e., novel) object from that sampled category and an object from a novel category in a choice phase. Mice preferentially explored the novel category object, taken as an index of category recognition, in this OCR task when a 30-min retention delay was used. Extensive preexposure to category exemplar objects also enhanced subsequent task performance across a longer (1-h) retention delay at which mice without preexposure did not demonstrate evidence for category recognition. Prechoice administration of the acetylcholine muscarinic receptor antagonist, scopolamine, disrupted OCR performance with or without preexposure, implicating acetylcholine in category recognition. The current study presents a valuable new rodent task for the study of the mechanistic basis of categorization-like processes and its potential relevance to common cognitive disorders. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Local-LDA: Open-Ended Learning of Latent Topics for 3D Object Recognition.

Service robots are expected to be more autonomous and work effectively in human-centric environments. This implies that robots should have special capabilities, such as learning from past experiences and real-time object category recognition. This paper proposes an open-ended 3D object recognition system which concurrently learns both the object categories and the statistical features for encoding objects. In particular, we propose an extension of Latent Dirichlet Allocation to learn structural semantic features (i.e., visual topics), from low-level feature co-occurrences, for each category independently. Moreover, topics in each category are discovered in an unsupervised fashion and are updated incrementally using new object views. In this way, the advantages of both the (hand-crafted) local features and the (learned) structural semantic features have been considered and combined in an efficient way. An extensive set of experiments has been performed to assess the performance of the proposed Local-LDA in terms of descriptiveness, scalability, and computation time. Experimental results show that the overall classification performance obtained with Local-LDA is clearly better than the best performances obtained with the state-of-the-art approaches. Moreover, the best scalability, in terms of number of learned categories, was obtained with the proposed Local-LDA approach, closely followed by a Bag-of-Words (BoW) approach. Concerning computation time, the best result was obtained with BoW, immediately followed by the Local-LDA approach.

The resonant brain: How attentive conscious seeing regulates action sequences that interact with attentive cognitive learning, recognition, and prediction

This article describes mechanistic links that exist in advanced brains between processes that regulate conscious attention, seeing, and knowing, and those that regulate looking and reaching. These mechanistic links arise from basic properties of brain design principles such as complementary computing, hierarchical resolution of uncertainty, and adaptive resonance. These principles require conscious states to mark perceptual and cognitive representations that are complete, context sensitive, and stable enough to control effective actions. Surface–shroud resonances support conscious seeing and action, whereas feature–category resonances support learning, recognition, and prediction of invariant object categories. Feedback interactions between cortical areas such as peristriate visual cortical areas V2, V3A, and V4, and the lateral intraparietal area (LIP) and inferior parietal sulcus (IPS) of the posterior parietal cortex (PPC) control sequences of saccadic eye movements that foveate salient features of attended objects and thereby drive invariant object category learning. Learned categories can, in turn, prime the objects and features that are attended and searched. These interactions coordinate processes of spatial and object attention, figure–ground separation, predictive remapping, invariant object category learning, and visual search. They create a foundation for learning to control motor-equivalent arm movement sequences, and for storing these sequences in cognitive working memories that can trigger the learning of cognitive plans with which to read out skilled movement sequences. Cognitive–emotional interactions that are regulated by reinforcement learning can then help to select the plans that control actions most likely to acquire valued goal objects in different situations. Many interdisciplinary psychological and neurobiological data about conscious and unconscious behaviors in normal individuals and clinical patients have been explained in terms of these concepts and mechanisms.

Innate colour preferences of a hawkmoth depend on visual context.

Solitary insects that feed on floral nectar must use innate knowledge to find their first flower. While innate preferences for flower colours are often described as fixed, species-specific traits, the nature and persistence of these preferences have been debated, particularly in relation to ontogenetic processes such as learning. Here we present evidence for a strong context-dependence of innate colour preferences in the crepuscular hawkmoth Manduca sexta. Contrary to expectations, our results show that innate colour biases shift with changes in the visual environment, namely illuminance and background. This finding reveals that innate responses might emerge from a contextual integration of sensory inputs involved in object class recognition rather than from the deterministic matching of such inputs with a fixed internal representation.

Supplemental Material for Development of an “Object Category Recognition” Task for Mice: Involvement of Muscarinic Acetylcholine Receptors

Supplemental Material for Development of an “Object Category Recognition” Task for Mice: Involvement of Muscarinic Acetylcholine Receptors

Using an Ensemble of Incrementally Fine-Tuned CNNs for Cross-Domain Object Category Recognition

When the training data is inadequate, it is difficult to train a deep Convolutional Neural Network (CNN) from scratch with randomized initial weights. Instead, it is common to train a source CNN model on a very large data set beforehand, and then use the learned source CNN model as an initialization to train a target CNN model. In deep learning realm, this procedure is called fine-tuning a CNN. This paper presents an experimental study on how to combine a collection of incrementally fine-tuned CNN models for cross-domain and multi-class object category recognition tasks. A group of fine-tuned CNN models is trained on the target data set by incrementally transferring parameters from a source CNN model trained on a large data set initially. The last two fully-connected (FC) layers of the source CNN model are eliminated, and two New FC layers are added to make the learned new CNN model suitable for the target task. Based on Caltech-101 and Office data sets, the experimental results demonstrate the effectiveness and good performance of the proposed methods. The proposed method is more suitable for the object recognition task when there is inadequate target training data.

Multi-Order Statistical Descriptors for Real-Time Face Recognition and Object Classification

We propose novel multi-order statistical descriptors which can be used for high speed object classification or face recognition from videos or image sets. We represent each gallery set with a global second-order statistic which captures correlated global variations in all feature directions as well as the common set structure. A lightweight descriptor is then constructed by efficiently compacting the second-order statistic using Cholesky decomposition. We then enrich the descriptor with the first-order statistic of the gallery set to further enhance the representation power. By projecting the descriptor into a low-dimensional discriminant subspace, we obtain further dimensionality reduction, while the discrimination power of the proposed representation is still preserved. Therefore, our method represents a complex image set by a single descriptor having significantly reduced dimensionality. We apply the proposed algorithm on image set and video-based face and periocular biometric identification, object category recognition, and hand gesture recognition. Experiments on six benchmark data sets validate that the proposed method achieves significantly better classification accuracy with lower computational complexity than the existing techniques. The proposed compact representations can be used for real-time object classification and face recognition in videos.

A study of the effect of noise and occlusion on the accuracy of convolutional neural networks applied to 3D object recognition

In this work, we carry out a study of the effect of adverse conditions, which characterize real-world scenes, on the accuracy of a Convolutional Neural Network applied to 3D object class recognition. Firstly, we discuss possible ways of representing 3D data to feed the network. In addition, we propose a set of representations to be tested. Those representations consist of a grid-like structure (fixed and adaptive) and a measure for the occupancy of each cell of the grid (binary and normalized point density). After that, we propose and implement a Convolutional Neural Network for 3D object recognition using Caffe. At last, we carry out an in-depth study of the performance of the network over a 3D CAD model dataset, the Princeton ModelNet project, synthetically simulating occlusions and noise models featured by common RGB-D sensors. The results show that the volumetric representations for 3D data play a key role on the recognition process and Convolutional Neural Network can be considerably robust to noise and occlusions if a proper representation is chosen.

Learning a deeply supervised multi-modal RGB-D embedding for semantic scene and object category recognition

Recognizing semantic category of objects and scenes captured using vision-based sensors is a challenging yet essential capability for mobile robots and UAVs to perform high-level tasks such as long-term autonomous navigation. However, extracting discriminative features from multi-modal inputs, such as RGB-D images, in a unified manner is non-trivial given the heterogeneous nature of the modalities. We propose a deep network which seeks to construct a joint and shared multi-modal representation through bilinearly combining the convolutional neural network (CNN) streams of the RGB and depth channels. This technique motivates bilateral transfer learning between the modalities by taking the outer product of each feature extractor output. Furthermore, we devise a technique for multi-scale feature abstraction using deeply supervised branches which are connected to all convolutional layers of the multi-stream CNN. We show that end-to-end learning of the network is feasible even with a limited amount of training data and the trained network generalizes across different datasets and applications. Experimental evaluations on benchmark RGB-D object and scene categorization datasets show that the proposed technique consistently outperforms state-of-the-art algorithms.