Adapting detectors to new datasets is needed in scenarios where a user has a specific dataset that contains novel classes or is recorded in a setting where a pretrained detector fails. While detectors based on Convolu...
详细信息
Adapting detectors to new datasets is needed in scenarios where a user has a specific dataset that contains novel classes or is recorded in a setting where a pretrained detector fails. While detectors based on Convolutional Neural Networks (CNNs) are state-of-the-art and nowadays publicly available, they suffer from bad generalization capabilities when applied on datasets that notably differ from the one they were trained on. Finetuning the detector is only possible if the dataset is large enough to not destroy the underlying feature representation. We propose a method where only a few prototypes are labeled for training in a semi-supervised manner. In particular, we separate the detection from the classification step to avoid impairing the bounding box proposal generation. Our trained prototype classification network provides labels to automatically source a large dataset containing 20 to 30 times more samples without further supervision, which we then use to train a more powerful network. We evaluate our method on a private vehicle dataset with six classes and show that evaluating on a previously unseen recording site we can gain an accuracy increase of 9% at same precision and recall levels. We further show that finetuning with as few as 25 labeled samples per class doubles accuracy compared to directly using pretrained features for nearest neighbor classification.
作者:
S KiruthikaV Vishnu PriyanAssistant professor
Department of Computer Science Engineering Sri Krishna College of Technology Coimbatore Student
Department of Computer Science Engineering Sri Krishna College of Technology Coimbatore
online groups offer clients with ways to overpower a few data hurdles and limitations, like the challenge to get self-governing data about movies and for the co-occurrence of positive and negative reactions within rev...
online groups offer clients with ways to overpower a few data hurdles and limitations, like the challenge to get self-governing data about movies and for the co-occurrence of positive and negative reactions within reviews. False reviews will disturb such choices because of misleading data, causing commercial disadvantages for the customers. Recognition of false opinions has thus expected very large concentration now days. But, many websites have only concentrated on handling the problematical comments and reviews. Our work tends to categorize people opinions into groups of true orfalse polarization by employing text feature analysis. In our work, our team analyze people opinions by implementing Sentiment Analysis techniques to recognize false opinions. Sentiment Analysis and text feature categorization techniques were used to a database containing people opinions. Furthermore, the estimation reviews acquired from reviewers could be categorized into good or bad opinions, that could be utilized by a customer to choose a movie. Further the proposed technique will graph based on the classification of true and fake reviews as the analysis of good and bad reviews for a product (movie). It will help us to predict the ratio of fake reviews to true reviews easily. To estimate the accomplishment of SA technique, this work has employed accurateness, exactness, recollection and F-degree as a performance rate.
Bitcoin is a rapidly growing but extremely risky cryptocurrency. It marks a watershed moment in the history of cash. These days, digital currency is preferred to actual money. Bitcoin has decentralized authority and p...
Bitcoin is a rapidly growing but extremely risky cryptocurrency. It marks a watershed moment in the history of cash. These days, digital currency is preferred to actual money. Bitcoin has decentralized authority and placed it in the hands of its users. Many people are joining the largest and most well-known Bitcoin mining pools as the risk of working alone is too great. In order to enhance their chances of creating the next block in the Bitcoins blockchain and decrease the mining reward volatility, users can band together to form Bitcoin pools. This tendency toward consolidation may also be seen in the rise of large-scale mining farms equipped with powerful mining resources and speedy processing capability. Because of the risk of a 51% assault, this pattern shows that Bitcoin’s pure, decentralized protocol is moving toward greater centralization in its distribution network. Not to be overlooked is the resulting centralization of the bitcoin network as a result of cloud wallets making it simple for new users to join. Because of the easily hackable nature of Bitcoin technologies, this could lead to a wide range of security vulnerabilities. The proposed approach uses normalization and filling missing values in preprocessing, PCA for feature Extraction and finally training the model using LSTM-DNN Models. The proposed approach outperforms other two models such as CNN and DNN.
Background Machine learning (ML) potential is not fully exploited in diagnostics and follow up of autoimmune inflammatory rheumatic disease (AIIRD). It is despite the broader use of ML in, e.g. imaging diagnostics. Th...
Background Machine learning (ML) potential is not fully exploited in diagnostics and follow up of autoimmune inflammatory rheumatic disease (AIIRD). It is despite the broader use of ML in, e.g. imaging diagnostics. The specific tools for AIIRD are lacking. Objectives This is an interim analysis of data from the first checkpoint of the proof-of-concept study on using accelerometer (ACC) data in follow up of patients with AIIRD. The main goal of the study was to investigate the value of single ACC data in the classification of arthritis activity status. Methods Subjects with AIIRD are enrolled in the study when they start of new treatment due to the disease activity. Several comorbidities, such as severe neurological and cardiological disorders, as wells as impaired mobility, are part of the protocol exclusion criteria. Volunteers without AIIRD who fulfil other inclusion/exclusion criteria are included as controls. The study was approved by local competent authorities and informed consent was given by all participants. We used data collected up till the first study checkpoint. This analysis covers nine patients with AIIRD (5 rheumatoid and four psoriasis arthritis) and 13 controls. Five patients had 3 study visits, 1 had 2 visits and 3 only one. Controls have only one visit per protocol. We analysed ACC data from 3 minutes of clapping using a home-brewed device based on Arduino nano 33 BLE with 6-axis MTU by Nordic Semiconductor. Data was divided into 6-second chunks that were found optimal in our prior study. We conducted binary classification between any/ no arthritis in any of the upper extremities. We used accuracy and area under curve (AUC) as an efficacy function derived from the receiver operating characteristic curve (ROC). We extracted 54 features from 3 ACC axis signal. The features encompassed, among others, Fourier's components, autoregression coefficients, median absolute deviation (MAD), variance, Fourier's entropy, etc. We built linear discriminant anal
This paper presents an advanced fuzzy C-means(FCM) clustering algorithm to overcome the weakness of the traditional FCM algorithm, including the instability of random selecting of initial center and the limitation of ...
详细信息
This paper presents an advanced fuzzy C-means(FCM) clustering algorithm to overcome the weakness of the traditional FCM algorithm, including the instability of random selecting of initial center and the limitation of the data separation or the size of clusters. The advanced FCM algorithm combines the distance with density and improves the objective function so that the performance of the algorithm can be improved. The experimental results show that the proposed FCM algorithm requires fewer iterations yet provides higher accuracy than the traditional FCM algorithm. The advanced algorithm is applied to the influence of stars' box-office data, and the classification accuracy of the first class stars achieves 92.625%.
In this paper we present four cases of minimal solutions for two-view relative pose estimation by exploiting the affine transformation between feature points and we demonstrate efficient solvers for these cases. It is...
详细信息
ISBN:
(数字)9781728171685
ISBN:
(纸本)9781728171692
In this paper we present four cases of minimal solutions for two-view relative pose estimation by exploiting the affine transformation between feature points and we demonstrate efficient solvers for these cases. It is shown, that under the planar motion assumption or with knowledge of a vertical direction, a single affine correspondence is sufficient to recover the relative camera pose. The four cases considered are two-view planar relative motion for calibrated cameras as a closed-form and a least-squares solution, a closed-form solution for unknown focal length and the case of a known vertical direction. These algorithms can be used efficiently for outlier detection within a RANSAC loop and for initial motion estimation. All the methods are evaluated on both synthetic data and real-world datasets from the KITTI benchmark. The experimental results demonstrate that our methods outperform comparable state-of-the-art methods in accuracy with the benefit of a reduced number of needed RANSAC iterations.
Diseases of the crop and insect pests are among the main causes of crop loss which is dangerous for agricultural production. Disease recognition is more difficult on the field because it has a complex background and d...
Diseases of the crop and insect pests are among the main causes of crop loss which is dangerous for agricultural production. Disease recognition is more difficult on the field because it has a complex background and different light intensity. Initial recognition and pests identification can significantly reduce the financial losses caused by pests. Using convolution neural networks, crop diseases can be automatically identified. The identification of a disease is often based on signs such as lesions or spots found in various slices of a plant. The size, color and the amount of these spots can define in great detail the disease that has killed the crop. Public data set is used as a data set. Experimental environment Model works on is a biodiverse environment.
Today in every industry weather, it is ISP, IT products, social network or mobile services there is the problem of customer churn (Customers changing their services from one service provider to another). However, in t...
详细信息
Today in every industry weather, it is ISP, IT products, social network or mobile services there is the problem of customer churn (Customers changing their services from one service provider to another). However, in telecommunication the customers churning very frequently. As the market in telecom is fiercely competitive, in that case, companies proactively have to determine the customers churn by analyzing their behavior and try to put effort and money in retaining the customers. In this proposed model, two machine-learning techniques were used for predicting customer churn Logistic regression and Logit Boost. Experiment was carried out in the WEKA Machine-learning tool, along with a real database from an American company Orange. The result were shown in different evaluation measures.
Pulmonary Tuberculosis (TB) one of the transmissible diseases, which is one of the top ten causes of death worldwide. The need to strengthen the treatment and screening in TB affected countries. In this paper, a syste...
Pulmonary Tuberculosis (TB) one of the transmissible diseases, which is one of the top ten causes of death worldwide. The need to strengthen the treatment and screening in TB affected countries. In this paper, a systematic review is carried on deep learning-based computer-aided diagnostic (CAD) systems that are used to analyze chest X-rays for diagnosing pulmonary tuberculosis (TB). Deep learning has recently become one of the most successful techniques, particularly in the analysis of medical images. In Deep learning Convolutional Neural Networks (CNNs) are widely used for TB detection. A CNN model is commonly comprised of convolutional layers, sub-sampling / pooling layers, and fully connected layers. This paper also presents a comprehensive survey on the CNN models for the detection of TB. The progression of computer-aided diagnostic (CAD) systems has sped up the early diagnosis of TB.
Virtual Intelligence can be otherwise known as digital intellect. VI is one of the developing technologies of this decade. It is the result of union of two technologies that are swiftly shooting up. They are Virtual R...
Virtual Intelligence can be otherwise known as digital intellect. VI is one of the developing technologies of this decade. It is the result of union of two technologies that are swiftly shooting up. They are Virtual Reality and Artificial Intelligence. While Artificial Intelligence will make machines react like an individual, VR will create an imaginary computer universe. This tech intends in transforming machines more like people. Using this VI technology, one can feel virtual world of their desire. This technology is utilized in many fields. Virtual Intelligence will integrate the mode of teaching. This will boost the gaming sector and will play a role in automation.
暂无评论