We present an effective system for integrating generative zero-shot classification modules into a YOLO-like dense detector to detect novel objects. Most double-stage-based novel object detection methods are achieved by refining the classification output branch but cannot be applied to a dense detector. Our system utilizes two paths to inject knowledge of novel objects into a dense detector. One involves injecting the class confidence for novel classes from a classifier trained on data synthesized via a dual-step generator. This generator learns a mapping function between two feature spaces, resulting in better classification performance. The second path involves re-training the detector head with feature maps synthesized on different intensity levels. This approach significantly increases the predicted objectness for novel objects, which is a major challenge for a dense detector. We also introduce a stop-and-reload mechanism during re-training for optimizing across head layers to better learn synthesized features. Our method relaxes the constraint on the detector head architecture in the previous method and has markedly enhanced performance on the MSCOCO dataset.

Aiming at the problem of non-line of sight (NLOS) signal recognition for Ultra Wide Band (UWB) positioning, we utilize the concepts of Neural Network Clustering and Neural Network Pattern Recognition. We propose a classification algorithm based on self-organizing feature mapping (SOM) neural network batch processing, and a recognition algorithm based on convolutional neural network (CNN). By assigning different weights to learning, training and testing parts in the data set of UWB location signals with given known patterns, a strong NLOS signal recognizer is trained to minimize the recognition error rate. Finally, the proposed NLOS signal recognition algorithm is verified using data sets from real scenarios. The test results show that the proposed algorithm can solve the problem of UWB NLOS signal recognition under strong signal interference. The simulation results illustrate that the proposed algorithm is significantly more effective compared with other algorithms.

In this paper, we propose an access control method with a secret key for object detection models for the first time so that unauthorized users without a secret key cannot benefit from the performance of trained models. The method enables us not only to provide a high detection performance to authorized users but to also degrade the performance for unauthorized users. The use of transformed images was proposed for the access control of image classification models, but these images cannot be used for object detection models due to performance degradation. Accordingly, in this paper, selected feature maps are encrypted with a secret key for training and testing models, instead of input images. In an experiment, the protected models allowed authorized users to obtain almost the same performance as that of non-protected models but also with robustness against unauthorized access without a key.

Recent researches on mobile robots show that convolutional neural network (CNN) has achieved impressive performance in visual place recognition especially for large-scale dynamic environment. However, CNN leads to the large space of image representation that cannot meet the real-time demand for robot navigation. Aiming at this problem, we evaluate the feature effectiveness of feature maps obtained from the layer of CNN by variance and propose a novel method that reserve salient feature maps and make adaptive binarization for them. Experimental results demonstrate the effectiveness and efficiency of our method. Compared with state of the art methods for visual place recognition, our method not only has no significant loss in precision, but also greatly reduces the space of image representation.

Maximum margin clustering (MMC) is a newly proposed clustering method that extends the large-margin computation of support vector machine (SVM) to unsupervised learning. Traditionally, MMC is formulated as a nonconvex integer programming problem which makes it difficult to solve. Several methods rely on reformulating and relaxing the nonconvex optimization problem as semidefinite programming (SDP) or second-order cone program (SOCP), which are computationally expensive and have difficulty handling large-scale data sets. In linear cases, by making use of the constrained concave-convex procedure (CCCP) and cutting plane algorithm, several MMC methods take linear time to converge to a local optimum, but in nonlinear cases, time complexity is still high. Since extreme learning machine (ELM) has achieved similar generalization performance at much faster learning speed than traditional SVM and LS-SVM, we propose an extreme maximum margin clustering (EMMC) algorithm based on ELM. It can perform well in nonlinear cases. Moreover, the kernel parameters of EMMC need not be tuned by means of random feature mappings. Experimental results on several real-world data sets show that EMMC performs better than traditional MMC methods, especially in handling large-scale data sets.

This paper presents a simple learning algorithm for network formation. The algorithm is based on self-organizing maps with growing cell structures and can adapt input data which correspond to nodes of the network. In basic numerical experiments, as a parameter is selected suitably, our algorithm can generate network having small-world-like structure. Such network structure appears in some natural networks and has advantages in practical systems.

Finite-state vector quantization (FSVQ) is a well-known block encoding technique for digital image compression at low bit rate application. In this paper, an improved feature map finite-state vector quantization (IFMFSVQ) algorithm using three-sided side-match prediction is proposed for image coding. The new three-sided side-match improves the prediction quality of input blocks. Precoded blocks are used to alleviate the error propagation of side-match. An edge threshold is used to classify the blocks into nonedge or edge blocks to improve bit rate performance. Furthermore, an adaptive method is also obtained. Experimental results reveal that the new IFMFSVQ reduces bit rate significantly maintaining the same subjective quality, as compared to the basic FMFSVQ method.

Recently, feature maps have been applied to various problem domains. The success of some of these applications critically depends on whether feature maps are topologically ordered. Several different approaches have been proposed to improve the conventional self-organizing feature map (SOM) algorithm. However, these approaches do not guarantee that a topologically-ordered feature map can be formed at the end of a simulation. Therefore, the trial-and-error procedure still dominates the procedure of forming feature maps. In this paper, we propose a healing mechanism to repair feature maps that are not well topologically ordered. The healed map is then further fine-tuned by the conventional SOM algorithm with a small learning rate and a small-sized neighborhood set so as to improve the accuracy of the map. Two data sets were tested to illustrate the performance of the proposed method.

There have been many developments on neural network research, and ability of a multi-layered network for classification of multi-spectral image data has been studied. We can classify non-Gaussian distributed data using the neural network trained by a back-propagation method (BPM) because it is independent of noise conditions. The BPM is a supervised classifier, so that we can get a high classification accuracy by using the method, so long as we can choose the good training data set. However, the multi-spectral data have many kinds of category information in a pixel because of its pixel resolution of the sensor. The data should be separated in many clusters even if they belong to a same class. Therefore, it is difficult to choose the good training data set which extract the characteristics of the class. Up to now, the researchers have chosen the training data set by random sampling from the input data. To overcome the problem, a hybrid pattern classification system using BPM and Kohonens feature mapping (KFM) has been proposed recently. The system performed choosing the training data set from the result of rough classification using KFM. However, how the remotely sensed data had been influenced by the KFM has not been demonstrated quantitatively. In this paper, we propose a new approach using the competitive weight vectors as the training data set, because we consider that a competitive unit represents a small cluster of the input patterns. The approach makes the training data set choice work easier than the usual one, because the KFM can automatically self-organize a topological relation among the target image patterns on a competitive plane. We demonstrate that the representative of the competitive units by principal component analysis (PCA). We also illustrate that the approach improves the classification accuracy by applying it on the classification of the real remotely sensed data.

In this paper, I investigate a property of self-organizing feature map (SOFM) in terms of reference vector density q(x) when probability density function of input signal fed into SOFM is p(x). Difficulty of general analysis on this property is briefly discussed. Then, I employ an assumption (conformal map assumption) to evaluate this property, and it is shown that for equilibrium state, q(x)p(x)s holds. By giving Lyapunov functioin for time evolution of reference vector density q(x) in SOFM, the equilibrium state is proved to be stable in terms of distribution. Comparison of the result with one which is based on different assumption reveals that there is no unique result of a simple form, such as conjectured by Kohonen. However, as there are cases in which these assumptions hold, these results suggest that we can consider a range of the property of SOFM. On the basis of it, we make comparison on this property between SOFM and fundamental adaptive vector quantization algorithm, in terms of the exponent s of the relation q(x)p(x)s. Difference on this property between SOFM and fundamental adaptive vector quantization algorithm, and propriety of mean squared quantization error for a performance measure of SOFM, are discussed.

The self-organizing feature map is one of the most widely used neural network paradigm based on unsupervised competitive learning. However, the learning algorithm introduced by Kohonen is very slow when the size of the map is large. The slowness is caused by the search for large map in each training steps of the learning. In this paper, a fast learning algorithm based on incremental ordering is proposed. The new learning starts with only a few units evenly distributed on a large topological feature map, and gradually increases the number of units until it covers the entire map. In middle phases of the learning, some units are well-ordered and others are not, while all units are weekly-ordered in Kohonen learning. The ordered units, during the learning, help to accelerate the search speed of the algorithm and accelerate the movements of the remaining unordered units to their topological locations. It is shown by theoretical analysis as well as experimental analysis that the proposed learning algorithm reduces the training time from O(M2) to O(log M) for M by M map without any additional working space, while preserving the ordering properties of the Kohonen learning algorithm.