Searcharxiv⌕ Search

arXiv subjects

Binbin Xu

Publications and source records attributed to Binbin Xu.

50 records · Page 3Linked to original sources
arXiv2022

Visual-Inertial Multi-Instance Dynamic SLAM with Object-level Relocalisation

In this paper, we present a tightly-coupled visual-inertial object-level multi-instance dynamic SLAM system. Even in extremely dynamic scenes, it can robustly optimise for the camera pose, velocity, IMU biases and build a dense 3D reconstruction object-level map of the environment. Our system can robustly track and reconstruct the geometries of arbitrary objects, their semantics and motion by incrementally fusing associated colour, depth, semantic, and foreground object probabilities into each object model thanks to its robust sensor and object tracking. In addition, when an object is lost or moved outside the camera field of view, our system can reliably recover its pose upon re-observation. We demonstrate the robustness and accuracy of our method by quantitatively and qualitatively testing it in real-world data sequences.

cs.RO↗
arXiv2022

Towards a Data-Driven Requirements Engineering Approach: Automatic Analysis of User Reviews

We are concerned by Data Driven Requirements Engineering, and in particular the consideration of user's reviews. These online reviews are a rich source of information for extracting new needs and improvement requests. In this work, we provide an automated analysis using CamemBERT, which is a state-of-the-art language model in French. We created a multi-label classification dataset of 6000 user reviews from three applications in the Health & Fitness field. The results are encouraging and suggest that it's possible to identify automatically the reviews concerning requests for new features. Dataset is available at: https://github.com/Jl-wei/APIA2022-French-user-reviews-classification-dataset.

cs.CL↗
arXiv2021

Identity-Disentangled Neural Deformation Model for Dynamic Meshes

Neural shape models can represent complex 3D shapes with a compact latent space. When applied to dynamically deforming shapes such as the human hands, however, they would need to preserve temporal coherence of the deformation as well as the intrinsic identity of the subject. These properties are difficult to regularize with manually designed loss functions. In this paper, we learn a neural deformation model that disentangles the identity-induced shape variations from pose-dependent deformations using implicit neural functions. We perform template-free unsupervised learning on 3D scans without explicit mesh correspondence or semantic correspondences of shapes across subjects. We can then apply the learned model to reconstruct partial dynamic 4D scans of novel subjects performing unseen actions. We propose two methods to integrate global pose alignment with our neural deformation model. Experiments demonstrate the efficacy of our method in the disentanglement of identities and pose. Our method also outperforms traditional skeleton-driven models in reconstructing surface details such as palm prints or tendons without limitations from a fixed template.

cs.CV↗
arXiv2021

Improving Patent Mining and Relevance Classification using Transformers

Patent analysis and mining are time-consuming and costly processes for companies, but nevertheless essential if they are willing to remain competitive. To face the overload induced by numerous patents, the idea is to automatically filter them, bringing only few to read to experts. This paper reports a successful application of fine-tuning and retraining on pre-trained deep Natural Language Processing models on patent classification. The solution that we propose combines several state-of-the-art treatments to achieve our goal - decrease the workload while preserving recall and precision metrics.

cs.CL↗
arXiv2021

A Benchmarking on Cloud based Speech-To-Text Services for French Speech and Background Noise Effect

This study presents a large scale benchmarking on cloud based Speech-To-Text systems: {Google Cloud Speech-To-Text}, {Microsoft Azure Cognitive Services}, {Amazon Transcribe}, {IBM Watson Speech to Text}. For each systems, 40158 clean and noisy speech files about 101 hours are tested. Effect of background noise on STT quality is also evaluated with 5 different Signal-to-noise ratios from 40dB to 0dB. Results showed that {Microsoft Azure} provided lowest transcription error rate $9.09\%$ on clean speech, with high robustness to noisy environment. {Google Cloud} and {Amazon Transcribe} gave similar performance, but the latter is very limited for time-constraint usage. Though {IBM Watson} could work correctly in quiet conditions, it is highly sensible to noisy speech which could strongly limit its application in real life situations.

cs.CL↗
arXiv2021

Pre-training A Neural Language Model Improves The Sample Efficiency of an Emergency Room Classification Model

To build a French national electronic injury surveillance system based on emergency room visits, we aim to develop a coding system to classify their causes from clinical notes in free-text. Supervised learning techniques have shown good results in this area but require a large amount of expert annotated dataset which is time consuming and costly to obtain. We hypothesize that the Natural Language Processing Transformer model incorporating a generative self-supervised pre-training step can significantly reduce the required number of annotated samples for supervised fine-tuning. In this preliminary study, we test our hypothesis in the simplified problem of predicting whether a visit is the consequence of a traumatic event or not from free-text clinical notes. Using fully re-trained GPT-2 models (without OpenAI pre-trained weights), we assess the gain of applying a self-supervised pre-training phase with unlabeled notes prior to the supervised learning task. Results show that the number of data required to achieve a ginve level of performance (AUC>0.95) was reduced by a factor of 10 when applying pre-training. Namely, for 16 times more data, the fully-supervised model achieved an improvement <1% in AUC. To conclude, it is possible to adapt a multi-purpose neural language model such as the GPT-2 to create a powerful tool for classification of free-text notes with only a small number of labeled samples.

cs.CL↗
arXiv2020

Deep Probabilistic Feature-metric Tracking

Dense image alignment from RGB-D images remains a critical issue for real-world applications, especially under challenging lighting conditions and in a wide baseline setting. In this paper, we propose a new framework to learn a pixel-wise deep feature map and a deep feature-metric uncertainty map predicted by a Convolutional Neural Network (CNN), which together formulate a deep probabilistic feature-metric residual of the two-view constraint that can be minimised using Gauss-Newton in a coarse-to-fine optimisation framework. Furthermore, our network predicts a deep initial pose for faster and more reliable convergence. The optimisation steps are differentiable and unrolled to train in an end-to-end fashion. Due to its probabilistic essence, our approach can easily couple with other residuals, where we show a combination with ICP. Experimental results demonstrate state-of-the-art performances on the TUM RGB-D dataset and the 3D rigid object tracking dataset. We further demonstrate our method's robustness and convergence qualitatively.

cs.CV↗
arXiv2020

A note on generalized $q$-difference equations for general Al-Salam--Carlitz polynomials

In this paper, we deduce the generalized $q$-difference equations for general Al-Salam--Carlitz polynomials and generalize Arjika's recently results [$q$-difference equation for homogeneous $q$-difference operators and their applications, J. Differ. Equ. Appl. {\bf 26}, 987--999 (2020)]. In addition, we obtain transformational identities by the method of $q$-difference equation. Moreover, we deduce $U(n+1)$ type generating functions and Ramanujan's integrals involving general Al-Salam--Carlitz polynomials by $q$-difference equation.

math.CO↗
arXiv2020

Bowditch's Q-conditions and Minsky's primitive stability

For the action of the outer automorphism group of the rank two free group on the corresponding variety of PSL(2,C) characters, two domains of discontinuity have been known to exist that are strictly larger than the set of Schottky characters. One is introduced by Bowditch in 1998 (followed by Tan, Wong and Zhang in 2008) and the other by Minsky in 2013. We prove that these two domains are equal. We then show that they are contained in the set of characters having what we call the bounded intersection property.

math.GT↗
arXiv2019

MID-Fusion: Octree-based Object-Level Multi-Instance Dynamic SLAM

We propose a new multi-instance dynamic RGB-D SLAM system using an object-level octree-based volumetric representation. It can provide robust camera tracking in dynamic environments and at the same time, continuously estimate geometric, semantic, and motion properties for arbitrary objects in the scene. For each incoming frame, we perform instance segmentation to detect objects and refine mask boundaries using geometric and motion information. Meanwhile, we estimate the pose of each existing moving object using an object-oriented tracking method and robustly track the camera pose against the static scene. Based on the estimated camera pose and object poses, we associate segmented masks with existing models and incrementally fuse corresponding colour, depth, semantic, and foreground object probabilities into each object model. In contrast to existing approaches, our system is the first system to generate an object-level dynamic volumetric map from a single RGB-D camera, which can be used directly for robotic tasks. Our method can run at 2-3 Hz on a CPU, excluding the instance segmentation part. We demonstrate its effectiveness by quantitatively and qualitatively testing it on both synthetic and real-world sequences.

cs.RO↗
arXiv2018

Carrier graphs for representations of the rank two free group into isometries of hyperbolic three space

Carrier graphs were first introduced for closed hyperbolic 3-manifolds by White. In this paper, we first generalize this definition to carrier graphs for representations of a rank two free group into the isometry group of hyperbolic three space. Then we prove the existence and the finiteness of minimal carrier graphs for those representations which are discrete, faithful and geometrically finite, and more generally, those that satisfy certain finiteness conditions first introduced by Bowditch.

math.GT↗
arXiv2016

Incompleteness of pressure metric on Teichmüller space of a bordered surface

We prove that the pressure metric on the Teichmüller space of a bordered surface is incomplete and its partial completion can be given by the moduli space of metric graphs for a fat graph associated to the same bordered surface equipped with pressure metric. As a corollary, we show that the pressure metric is not a constant multiple of the Weil-Petersson metric which is different from the closed surface case.

math.DS↗
arXiv2016

Nonlinear trend removal should be carefully performed in heart rate variability analysis

$\bullet$ Background : In Heart rate variability analysis, the rate-rate time series suffer often from aperiodic non-stationarity, presence of ectopic beats etc. It would be hard to extract helpful information from the original signals. 10 $\bullet$ Problem : Trend removal methods are commonly practiced to reduce the influence of the low frequency and aperiodic non-stationary in RR data. This can unfortunately affect the signal and make the analysis on detrended data less appropriate. $\bullet$ Objective : Investigate the detrending effect (linear \& nonlinear) in temporal / nonliear analysis of heart rate variability of long-term RR data (in normal sinus rhythm, atrial fibrillation, 15 congestive heart failure and ventricular premature arrhythmia conditions). $\bullet$ Methods : Temporal method : standard measure SDNN; Nonlinear methods : multi-scale Fractal Dimension (FD), Detrended Fluctuation Analysis (DFA) \& Sample Entropy (Sam-pEn) analysis. $\bullet$ Results : The linear detrending affects little the global characteristics of the RR data, either 20 in temporal analysis or in nonlinear complexity analysis. After linear detrending, the SDNNs are just slightly shifted and all distributions are well preserved. The cross-scale complexity remained almost the same as the ones for original RR data or correlated. Nonlinear detrending changed not only the SDNNs distribution, but also the order among different types of RR data. After this processing, the SDNN became indistinguishable be-25 tween SDNN for normal sinus rhythm and ventricular premature beats. Different RR data has different complexity signature. Nonlinear detrending made the all RR data to be similar , in terms of complexity. It is thus impossible to distinguish them. The FD showed that nonlinearly detrended RR data has a dimension close to 2, the exponent from DFA is close to zero and SampEn is larger than 1.5 -- these complexity values are very close to those for 30 random signal. $\bullet$ Conclusions : Pre-processing by linear detrending can be performed on RR data, which has little influence on the corresponding analysis. Nonlinear detrending could be harmful and it is not advisable to use this type of pre-processing. Exceptions do exist, but only combined with other appropriate techniques to avoid complete change of the signal's intrinsic dynamics. 35 Keywords $\bullet$ heart rate variability $\bullet$ linear / nonlinear detrending $\bullet$ complexity analysis $\bullet$ mul-tiscale analysis $\bullet$ detrended fluctuation analysis $\bullet$ fractal dimension $\bullet$ sample entropy;

q-bio.QM↗