The Oxford Radar RobotCar dataset has recently become popular in evaluating LiDAR-based methods for place recognition. The Radar dataset is preferred over the original Oxford RobotCar dataset since it has better LiDAR sensors and location ground truth is available for all sequences. However, it turns out that the Radar dataset has serious issues with its ground truth and therefore experimental findings with this dataset can be misleading. We demonstrate how easily this can happen, by varying only the gallery sequence and keeping the training and test sequences fixed. Results of this experiment strongly indicate that the gallery selection is an important consideration for place recognition. However, the finding is a mistake and the difference between galleries can be explained by systematic errors in the ground truth. In this work, we propose a revised benchmark for LiDAR-based place recognition with the Oxford Radar RobotCar dataset. The benchmark includes fixed gallery, training and test sequences, corrected ground truth, and a strong baseline method. All data and code will be made publicly available to facilitate fair method comparison and development.

Image classification is an essential challenge for many types of autonomous and smart systems. With advances in Convolutional Neural Networks (CNNs), the accuracy of image classification systems has been dramatically improved. However, due to the escalating complexity of state-of-the-art CNN solutions, significant challenges arise in implementing real-time image classification applications on resource-constrained platforms. The framework of elastic neural networks has been proposed to address trade-offs between classification accuracy and real-time performance by leveraging intermediate early-exits placed in deep CNNs and allowing systems to switch among multiple candidate outputs, while switching off inference layers that are not used by the selected output. In this paper, we propose a novel approach for configuring early-exit points when converting a deep CNN into an elastic neural network. The proposed approach is designed to systematically optimize the quality and diversity of the alternative CNN operating points that are provided by the derived elastic networks. We demonstrate the utility of the proposed elastic neural network approach on the CIFAR-100 dataset.

BACKGROUND AND OBJECTIVE:Emergency Department (ED) overcrowding is a chronic international issue that is associated with adverse treatment outcomes. Accurate forecasts of future service demand would enable intelligent resource allocation that could alleviate the problem. There has been continued academic interest in ED forecasting but the number of used explanatory variables has been low, limited mainly to calendar and weather variables. In this study we investigate whether predictive accuracy of next day arrivals could be enhanced using high number of potentially relevant explanatory variables and document two feature selection processes that aim to identify which subset of variables is associated with number of next day arrivals. Performance of such predictions over longer horizons is also shown.METHODS:We extracted numbers of total daily arrivals from Tampere University Hospital ED between the time period of June 1, 2015 and June 19, 2019. 158 potential explanatory variables were collected from multiple data sources consisting not only of weather and calendar variables but also an extensive list of local public events, numbers of website visits to two hospital domains, numbers of available hospital beds in 33 local hospitals or health centres and Google trends searches for the ED. We used two feature selection processes: Simulated Annealing (SA) and Floating Search (FS) with Recursive Least Squares (RLS) and Least Mean Squares (LMS). Performance of these approaches was compared against autoregressive integrated moving average (ARIMA), regression with ARIMA errors (ARIMAX) and Random Forest (RF). Mean Absolute Percentage Error (MAPE) was used as the main error metric.RESULTS:Calendar variables, load of secondary care facilities and local public events were dominant in the identified predictive features. RLS-SA and RLS-FA provided slightly better accuracy compared ARIMA. ARIMAX was the most accurate model but the difference between RLS-SA and RLS-FA was not statistically significant.CONCLUSIONS:Our study provides new insight into potential underlying factors associated with number of next day presentations. It also suggests that predictive accuracy of next day arrivals can be increased using high-dimensional feature selection approach when compared to both univariate and nonfiltered high-dimensional approach. Performance over multiple horizons was similar with a gradual decline for longer horizons. However, outperforming ARIMAX remains a challenge when working with daily data. Future work should focus on enhancing the feature selection mechanism, investigating its applicability to other domains and in identifying other potentially relevant explanatory variables.

In this paper, we provide a comprehensive study on evaluating two state-of-the-art deep metric learning methods for visual place recognition. Visual place recognition is an essential component in the visual localization and the vision-based navigation where it provides an initial coarse location. It is used in variety of autonomous navigation technologies, including autonomous vehicles, drones and computer vision systems. We study recent visual place recognition and image retrieval methods and utilize them to conduct extensive and comprehensive experiments on two diverse and large long-term indoor and outdoor robot navigation datasets, e.g., COLD and Oxford Radar RobotCar along with ablation studies on the crucial parameters of the deep architectures. Our comprehensive results indicate that the methods can achieve 5 m of outdoor and 50 cm of indoor place recognition accuracy with high recall rate of 80 %.

In this work, we study the long-term performance of visual place recognition in urban outdoor environment. A long-term benchmark is constructed from the Oxford RobotCar dataset. It contains sequences of the same route traversed over a period of approx. 500 days. We carefully selected three gallery sequences, one training sequence and 15 query sequences that cover different seasons, times of day and weather. The RobotCar sequences from the first half year have several problems, for example, only partial routes and inaccurate location data. We circumvent these problems by reversing the time. In the benchmark dataset the gallery and training images are the latest and the query sequences go gradually back in time. Our experiments provide the following findings. 1) the selected gallery sequence has strong impact on performance, and 2) additional training sequences help to mitigate differences between the gallery sequences. In addition, results indicate that 3) there is a long-term trend of performance degradation over time. The degradation can be quantified as about 6 percentage points per 100 days and, therefore, the initial performance of 40% eventually drops below 20% at the end.

Place recognition is one of the main challenges in localization, mapping and navigation tasks of self-driving vehicles under various perceptual conditions, including appearance and viewpoint variations. In this paper, we provide a comprehensive study on the utility of fine-tuned Deep Convolutional Neural Network (DCNN) with three MAC, SpoC and GeM pooling layers to learn global image representation for place recognition in an end-to-end manner using three different sensor data modalities: (1) only RGB images; (2) only intensity or only depth 3D LiDAR point clouds projected into 2D images and (3) early fusion of RGB images and LiDAR point clouds (both intensity and depth) to form a unified global descriptor to leverage robust features of both modalities. The experimental results on a diverse and large long-term Oxford Radar RobotCar dataset illustrate an achievement of 5 m outdoor place recognition accuracy with high recall rate of 90 % using early fusion of RGB and LiDAR sensor data modalities when fine-tuned network with GeM pooling layer is utilized.

Manual annotation of bounding boxes for object detection in digital images is tedious, and time and resource consuming. In this paper, we propose a semi-automatic method for efficient bounding box annotation. The method trains the object detector iteratively on small batches of labeled images and learns to propose bounding boxes for the next batch, after which the human annotator only needs to correct possible errors. We propose an experimental setup for simulating the human actions and use it for comparing different iteration strategies, such as the order in which the data is presented to the annotator. We experiment on our method with three datasets and show that it can reduce the human annotation effort significantly, saving up to 75% of total manual annotation work.

In this paper, we propose a simple yet effective method to deal with the violation of the Closed-World Assumption for a classifier. Previous works tend to apply a threshold either on the classification scores or the loss function to reject the inputs that violate the assumption. However, these methods cannot achieve the low False Positive Ratio (FPR) required in safety applications. The proposed method is a rejection option based on hypothesis testing with probabilistic networks. With probabilistic networks, it is possible to estimate the distribution of outcomes instead of a single output. By utilizing Z-test over the mean and standard deviation for each class, the proposed method can estimate the statistical significance of the network certainty and reject uncertain outputs. The proposed method was experimented on with different configurations of the COCO and CIFAR datasets. The performance of the proposed method is compared with the Softmax Response, which is a known top-performing method. It is shown that the proposed method can achieve a broader range of operation and cover a lower FPR than the alternative.

Supervised object detection has been proven to be successful in many benchmark datasets achieving human-level performances. However, acquiring a large amount of labeled image samples for supervised detection training is tedious, time-consuming, and costly. In this paper, we propose an efficient image selection approach that samples the most informative images from the unlabeled dataset and utilizes human-machine collaboration in an iterative train-annotate loop. Image features are extracted by the CNN network followed by the similarity score calculation, Euclidean distance. Unlabeled images are then sampled into different approaches based on the similarity score. The proposed approach is straightforward, simple and sampling takes place prior to the network training. Experiments on datasets show that our method can reduce up to 80% of manual annotation workload, compared to full manual labeling setting, and performs better than random sampling.

In this study, we analyze model inversion attacks with only two assumptions: feature vectors of user data are known, and a black-box API for inference is provided. On the one hand, limitations of existing studies are addressed by opting for a more practical setting. Experiments have been conducted on state-of-the-art models in person re-identification, and two attack scenarios (i.e., recognizing auxiliary attributes and reconstructing user data) are investigated. Results show that an adversary could successfully infer sensitive information even under severe constraints. On the other hand, it is advisable to encrypt feature vectors, especially for a machine learning model in production. As an alternative to traditional encryption methods such as AES, a simple yet effective method termed ShuffleBits is presented. More specifically, the binary sequence of each floating-point number gets shuffled. Deployed using the one-time pad scheme, it serves as a plug-and-play module that is applicable to any neural network, and the resulting model directly outputs deep features in encrypted form. Source code is publicly available at https://github.com/nixingyang/ShuffleBits.

Age estimation is an essential challenge in computer vision. With the advances of convolutional neural networks, the performance of age estimation has been dramatically improved. Existing approaches usually treat age estimation as a classification problem. However, the age labels are ambiguous, thus make the classification task difficult. In this paper, we propose a simple yet effective approach for age estimation, which improves the performance compared to classification-based methods. The method combines four classification losses and one regression loss representing different class granularities together, and we name it as Age-Granularity-Net. We validate the Age-Granularity-Net framework on the CVPR Chalearn 2016 dataset, and extensive experiments show that the proposed approach can reduce the prediction error compared to any individual loss. The source code link is https://github.com/yipersevere/age-estimation.

We introduce an adaptive L-2 regularization mechanism in the setting of person re-identification. In the literature, it is common practice to utilize hand-picked regularization factors which remain constant throughout the training procedure. Unlike existing approaches, the regularization factors in our proposed method are updated adaptively through backpropagation. This is achieved by incorporating trainable scalar variables as the regularization factors, which are further fed into a scaled hard sigmoid function. Extensive experiments on the Market-1501, DukeMTMC-reID and MSMT17 datasets validate the effectiveness of our framework. Most notably, we obtain stateof-the-art performance on MSMT17, which is the largest dataset for person re-identification. Source code is publicly available at https://github.com/nixingyang/AdaptiveL2Regularization.

Label noise is a primary point of interest for safety concerns in previous works as it affects the robustness of a machine learning system by a considerable amount. This paper studies the sensitivity of object detection loss functions to label noise in bounding box detection tasks. Although label noise has been widely studied in the classification context, less attention is paid to its effect on object detection. We characterize different types of label noise and concentrate on the most common type of annotation error, which is missing labels. We simulate missing labels by deliberately removing bounding boxes at training time and study its effect on different deep learning object detection architectures and their loss functions. Our primary focus is on comparing two particular loss functions: cross-entropy loss and focal loss. We also experiment on the effect of different focal loss hyperparameter values with varying amounts of noise in the datasets and discover that even up to 50% missing labels can be tolerated with an appropriate selection of hyperparameters. The results suggest that focal loss is more sensitive to label noise, but increasing the gamma value can boost its robustness.

In this work, loop-closure detection from LiDAR scans is defined as an image re-identification problem. Reidentification is performed by computing Euclidean distances of a query scan to a gallery set of previous scans. The distances are computed in a feature embedding space where the scans are mapped by a convolutional neural network (CNN). The network is trained using the triplet loss training strategy. In our experiments we compare different backbone networks, variants of the triplet loss and generic and LiDAR specific data augmentation techniques. With a realistic indoor dataset the best architecture obtains the mean average precision (mAP) above 0.94.

Facial analysis is an active research area in computer vision, with many practical applications. Most of the existing studies focus on addressing one specific task and maximizing its performance. For a complete facial analysis system, one needs to solve these tasks efficiently to ensure a smooth experience. In this work, we present a system-level design of a real-time facial analysis system. With a collection of deep neural networks for object detection, classification, and regression, the system recognizes age, gender, facial expression, and facial similarity for each person that appears in the camera view. We investigate the parallelization and interplay of individual tasks. Results on common off-the-shelf architecture show that the system's accuracy is comparable to the state-of-the-art methods, and the recognition speed satisfies real-time requirements. Moreover, we propose a multitask network for jointly predicting the first three attributes, i.e., age, gender, and facial expression. Source code and trained models are available at https://github.com/mahehu/TUT-live-age-estimator.

We compare a state-of-the-art deep image retrieval and a deep place recognition method for place recognition using LiDAR data. Place recognition aims to detect previously visited locations and thus provides an important tool for navigation, mapping, and localisation. Experimental comparisons are conducted using challenging outdoor and indoor datasets, Oxford Radar RobotCar and COLD, in the "long-term" setting where the test conditions differ substantially from the training and gallery data. Based on our results the image retrieval methods using LiDAR depth images can achieve accurate localization (the single best match recall 80%) within 5.00 m in urban outdoors. In office indoors the comparable accuracy is 50 cm but is more sensitive to changes in the environment.

This paper studies vehicle attribute recognition by appearance. In the literature, image-based target recognition has been extensively investigated in many use cases, such as facial recognition, but less so in the field of vehicle attribute recognition. We survey a number of algorithms that identify vehicle properties ranging from coarse-grained level (vehicle type) to fine-grained level (vehicle make and model). Moreover, we discuss two alternative approaches for these tasks, including straightforward classification and a more flexible metric learning method. Furthermore, we design a simulated real-world scenario for vehicle attribute recognition and present an experimental comparison of the two approaches.

In this paper we study the recognition of handwritten characters from data captured by a novel wearable electro-textile sensor panel. The data is collected sequentially, such that we record both the stroke order and the resulting bitmap. We propose a preprocessing pipeline that fuses the sequence and bitmap representations together. The data is collected from ten subjects containing altogether 7500 characters. We also propose a convolutional neural network architecture, whose novel upsampling structure enables successful use of conventional ImageNet pretrained networks, despite the small input size of only 10x10 pixels. The proposed model reaches 72\% accuracy in experimental tests, which can be considered good accuracy for this challenging dataset. Both the data and the model are released to the public.

The popularity of tools for analyzing Technical Debt, and particularly the popularity of SonarQube, is increasing rapidly. SonarQube proposes a set of coding rules, which represent something wrong in the code that will soon be reflected in a fault or will increase maintenance effort. However, our local companies were not confident in the usefulness of the rules proposed by SonarQube and contracted us to investigate the fault-proneness of these rules. In this work we aim at understanding which SonarQube rules are actually fault-prone and to understand which machine learning models can be adopted to accurately identify fault-prone rules. We designed and conducted an empirical study on 21 well-known mature open-source projects. We applied the SZZ algorithm to label the fault-inducing commits. We analyzed the fault-proneness by comparing the classification power of seven machine learning models. Among the 202 rules defined for Java by SonarQube, only 25 can be considered to have relatively low fault-proneness. Moreover, violations considered as “bugs” by SonarQube were generally not fault-prone and, consequently, the fault-prediction power of the model proposed by SonarQube is extremely low. The rules applied by SonarQube for calculating technical debt should be thoroughly investigated and their harmfulness needs to be further confirmed. Therefore, companies should carefully consider which rules they really need to apply, especially if their goal is to reduce fault-proneness.

This paper introduces a prototype of ClothFace technology, a battery-free textile-based handwriting recognition platform that includes an e-textile antenna and a 10 × 10 array of radio frequency identification (RFID) integrated circuits (ICs), each with a unique ID. Touching the textile platform surface creates an electrical connection from specific ICs to the antenna, which enables the connected ICs to be read with an external UHF (ultra-haigh frequency) RFID reader. In this paper, the platform is demonstrated to recognize handwritten numbers 0–9. The raw data collected by the platform are a sequence of IDs from the touched ICs. The system converts the data into bitmaps and their details are increased by interpolating between neighboring samples using the sequential information of IDs. These images of digits written on the platform can be classified, with enough accuracy for practical use, by deep learning. The recognition system was trained and tested with samples from six volunteers using the platform. The real-time number recognition ability of the ClothFace technology is demonstrated to work successfully with a very low error rate. The overall recognition accuracy of the platform is 94.6% and the accuracy for each digit is between 91.1% and 98.3%. As the solution is fully passive and gets all the needed energy from the external RFID reader, it enables a maintenance-free and cost-effective user interface that can be integrated into clothing and into textiles around us.