When building a unified vision system or gradually adding new capabilities to a system, the usual assumption is that training data for all tasks is always available. However, as the number of tasks grows, storing and retraining on such data becomes infeasible. A new problem arises where we add new capabilities to a Convolutional Neural Network (CNN), but the training data for its existing capabilities are unavailable. We propose our Learning without Forgetting method, which uses only new task data to train the network while preserving the original capabilities. Our method performs favorably compared to commonly used feature extraction and fine-tuning adaption techniques and performs similarly to multitask learning that uses original task data we assume unavailable. A more surprising observation is that Learning without Forgetting may be able to replace fine-tuning with similar old and new task datasets for improved new task performance.

Learning without Forgetting

In the near future, deep learning algorithms will be incorporated in several applications for assisting the human beings. The deep learning algorithms have the tendency to allow a computer to work on its assumption. Most of the deep learning algorithms mimic the human brain’s neuron connection to leverage an artificial intelligence to the computer system. This helps to improve the operational speed and accuracy on several critical tasks. This paper projects the blocks, which are required for the incorporation of deep learning based algorithm. Also, the paper attempts to deeply analyze the necessity of the preprocessing step over several deep learning based applications.

https://irojournals.com/iroiip/V3/I1/06.pdf

A Study to Find Facts Behind Preprocessing on Deep Learning Algorithms

The rapid growth of Internet users led to unwanted cyber issues, including cyberbullying, hate speech, and many more. This article deals with the problems of hate speech on Twitter. Hate speech appears to be an inflammatory kind of interaction process that uses misconceptions to express a hate ideology. The hate speech focuses on various protected aspects, including gender, religion, race, and disability. Owing to hate speech, sometimes unwanted crimes are going to happen as someone or a group of people get disheartened. Hence, it is essential to monitor user’s posts and filter the hate speech related post before it is spread. However, Twitter receives more than six hundred tweets per second and about 500 million tweets per day. Manually filtering any information from such a huge incoming traffic is almost impossible. Concerning to this aspect, an automated system is developed using the Deep Convolutional Neural Network (DCNN). The proposed DCNN model utilises the tweet text with GloVe embedding vector to capture the tweets’ semantics with the help of convolution operation and achieved the precision, recall and F1-score value as 0.97, 0.88, 0.92 respectively for the best case and outperformed the existing models.

/pdf/a-framework-for-hate-speech-detection-using-deep-3irdjzxtc4.pdf

A Framework for Hate Speech Detection Using Deep Convolutional Neural Network

In recent years, unethical behavior in the cyber-environment has been revealed. The presence of offensive language on social media platforms and automatic detection of such language is becoming a major challenge in modern society. The complexity of natural language constructs makes this task even more challenging. Until now, most of the research has focused on resource-rich languages like English. Roman Urdu and Urdu are two scripts of writing the Urdu language on social media. The Roman script uses the English language characters while the Urdu script uses Urdu language characters. Urdu and Hindi languages are similar with the only difference in their writing script but the Roman scripts of both languages are similar. This study is about the detection of offensive language from the user's comments presented in a resource-poor language Urdu. We propose the first offensive dataset of Urdu containing user-generated comments from social media. We use individual and combined n-grams techniques to extract features at character-level and word-level. We apply seventeen classifiers from seven machine learning techniques to detect offensive language from both Urdu and Roman Urdu text comments. Experiments show that the regression-based models using character n-grams show superior performance to process the Urdu language. Character-level tri-gram outperforms the other word and character n-grams. LogitBoost and SimpleLogistic outperform the other models and achieve 99.2% and 95.9% values of F-measure on Roman Urdu and Urdu datasets respectively. Our designed dataset is publically available on GitHub for future research.

/pdf/automatic-detection-of-offensive-language-for-urdu-and-roman-2pdisqxwe3.pdf

Automatic Detection of Offensive Language for Urdu and Roman Urdu

The invention relates to the field of a robot, in particular to an industrial robot control system based on visual positioning and a control method thereof. The control system provided by the invention is convenient to operate and has the advantages of simple structure and high efficiency and accuracy. In artificial teaching process, positioning motion control process and oriented motion control process of the control method provided by the invention, the positioning of luminescence mark points or the ends of an operating tool is performed in the coordinates ow, xw, yw and zy of a visual measuring device, thus the control system has no requirement on the motion control accuracy of the robot and the relative position relation of the coordinates of the tool and the coordinates of the robot, and the control method only requires linear motion realized by the rotation of axes from the first axis to the third axis and rotary motion realized by the rotation of axes of axes from the fourth axis to the sixth axis according to the theoretical model of the robot and, thereby fundamentally avoiding the extremely strict requirements on the accuracy of a robot model and the calibrating accuracy of the coordinates of the tool in conventional off-line programming teaching and meeting the operating requirements of a complicated track with high accuracy.

Industrial robot control system based on visual positioning and control method thereof

The rapid growth of electronic documents are causing problems like unstructured data that need more time and effort to search a relevant document. Text Document Classification (TDC) has a great significance in information processing and retrieval where unstructured documents are organized into pre-defined classes. Urdu is the most favorite research language in South Asian languages because of its complex morphology, unique features, and lack of linguistic resources like standard datasets. As compared to short text, like sentiment analysis, long text classification needs more time and effort because of large vocabulary, more noise, and redundant information. Machine Learning (ML) and Deep Learning (DL) models have been widely used in text processing. Despite the major limitations of ML models, like learn directed features, these are the favorite methods for Urdu TDC. To the best of our knowledge, it is the first study of Urdu TDC using DL model. In this paper, we design a large multi-purpose and multi-format dataset that contain more than ten thousand documents organize into six classes. We use Single-layer Multisize Filters Convolutional Neural Network (SMFCNN) for classification and compare its performance with sixteen ML baseline models on three imbalanced datasets of various sizes. Further, we analyze the effects of preprocessing methods on SMFCNN performance. SMFCNN outperformed the baseline classifiers and achieved 95.4%, 91.8%, and 93.3% scores of accuracy on medium, large and small size dataset respectively. The designed dataset would be publically and freely available in different formats for future research in Urdu text processing.

/pdf/document-level-text-classification-using-single-layer-4l8lnycyh6.pdf

Document-Level Text Classification Using Single-Layer Multisize Filters Convolutional Neural Network

Underwater acoustic classification is a challenging problem because of presence of high background noise and complex sound propagation patterns in the sea environment. Various algorithms proposed in last few years used own privately collected datasets for design and validation. Such data is not publicly available. To conduct research in this field, there is a dire need of publicly available dataset. To bridge this gap, we construct and present an underwater acoustic dataset, named DeepShip, which consists of 47 h and 4 min of real world underwater recordings of 265 different ships belong to four classes. The proposed dataset includes recording from throughout the year with different sea states and noise levels. The presented dataset will not only help to evaluate the performance of existing algorithms but it shall also benefit the research community in future. Using the proposed dataset, we also conducted a comprehensive study of various machine learning and deep learning algorithms on six time–frequency based extracted features. In addition, we propose a novel separable convolution based autoencoder network for better classification accuracy. Experiments results, which are compared based on classification accuracy, precision, recall, f1-score, and analyzed by using paired sampled statistical t-test, show that the proposed network achieves classification accuracy of 77.53% using CQT feature, which is better than as achieved by other methods.

DeepShip: An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification

The invention discloses a movement capturing method based on multiple binocular stereo vision. A movement video collecting device is constructed, and human movement video sequences from different orientations are collected by the movement video collecting device. Multiocular movement video sequences shot by a plurality of cameras are calibrated. Marked points matching and tracking of each binocular tracker is finished. Data fusion of three-dimensional tracking result of multiple binocular trackers is completed. The three-dimensional movement information of the marked points acquired by a multiocular fusion device is fed back to the binocular tracker to consummate binocular tracking. On the basis of binocular three-dimensional tracking realized by binocular vision, the invention fuses multiple groups of binocular three-dimensional movement data, resolves parameter acquiring problem of three-dimensional position, tracking, track fusion and the like for a plurality of marked points, increases number of traceable market points and enables the tracking effect to be comparable with three-dimensional movement acquiring device employing multi-infrared cameras for collecting.

Zheng Jiangbin

Papers

Document-Level Text Classification Using Single-Layer Multisize Filters Convolutional Neural Network

Automatic Detection of Offensive Language for Urdu and Roman Urdu

DeepShip: An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification

Method for capturing movement based on multiple binocular stereovision

A novel lifelong learning model based on cross domain knowledge extraction and transfer to classify underwater images