Greece

With image processing, robots acquired visual perception skills; enabling them to become autonomous. Since the emergence of Artificial Intelligence (AI), sophis-ticated tasks such as object identification have become possible through inferenc-ing Artificial Neural Networks (ANN). Be that as it may, Autonomous Mobile Robots (AMR) are Embedded Systems (ESs) with limited on-board resources. Thus, efficient techniques in ANN inferencing are required for real-time perfor-mance. This paper presents the process of optimizing ANNs inferencing using tensor-based optimization on embedded Graphical Processing Unit (GPU) with Computer Unified Device Architecture (CUDA) platform for parallel acceleration on ES. This research evaluates renowned network, namely, You-Only-Look-Once (YOLO), on NVIDIA Jetson TX2 System-On-Module (SOM). The find-ings of this paper display a significant improvement in inferencing speed in terms of Frames-Per-Second (FPS) up to 3.5 times the non-optimized inferencing speed. Furthermore, the current CUDA model and TensorRT optimization tech-niques are studied, comments are made on its implementation for inferencing, and improvements are proposed based on the results acquired. These findings will contribute to ES developers and industries will benefit from real-time perfor-mance inferencing for AMR automation solutions

AIAI 2020

Tensor-based CUDA Optimization for ANN Inferencing using Parallel Acceleration on Embedded GPU

nvidia jetson

tensorrt

embedded gpu

cuda

yolo

image processing

artificial neural networks

real-time

AIAI 2020 (16th International Conference on Artificial Intelligence Applications and Innovations) should be held in Halkidiki, Greece. Due to current Covid-19 situation the conference will be held virtually. Underline will host a repository of video talks and preserve them for the future. Videos will be released on Underline one day after the conference.

technical paper

Authorship attribution attempts to identify the authors behind texts and has important applications mainly in cyber-security, digital humanities and social media analytics. An especially challenging but very realistic scenario is cross-domain attribution where texts of known authorship (training set) differ from texts of disputed authorship (test set) in topic or genre. In this paper, we modify a successful authorship verification approach based on a multi-headed neural network language model and combine it with pre-trained language models. Based on experiments on a controlled corpus covering several text genres where topic and genre is specifically controlled, we demonstrate that the proposed approach achieves very promising results. We also demonstrate the crucial effect of the normalization corpus in cross-domain attribution.

Cross-domain Authorship Attribution Using Pre-trained Language Models

Everybody likes to make a good prediction, in particular,
when some sort of personal investment is involved in terms of finance, energy or time. The difficulty is to make a prediction that optimises the reward obtained from the original contribution; this is even more important when investments are the core service offered by a business or pension fund generated by monthly contributions. The complexity of finance is that the human predictor may have other interests or bias than the human investor, the trust between predictor and investor will never
be completely established as the investor will never know if the predictor has generated, intentionally or unintentionally, the optimum possible reward. This paper presents the Random Neural Network in recurrent configuration that makes predictions on time series data, specifically, prices. The biological model inspired by the brain structure and neural interconnections makes predictions entirely on previous data from the time series rather than predictions based on several uncorrelated inputs. The model is validated against the property, stock and Fintech market: 1) UK property prices, 2) stock markets indice prices, 3) cryptocurrency
prices. Experimental results show that the proposed method makes accurate predictions on different investment portfolios.

The Random Neural Network in Price Predictions

We investigated age related synaptic plasticity in thalamic reticular nucleus (TRN) as a part of visual information processing system in the brain. Simulation experiments were performed using a hierarchical spike timing neural network model in NEST simulator. The model consists of multiple layers starting with retinal photoreceptors through thalamic relay, primary visual cortex layers up to the lateral intraparietal cortex (LIP) responsible for decision making and preparation of motor response. All synaptic inter- and intra-layer connections of our model are structured according to the literature information. The present work extends the model with spike timing dependent plastic (STDP) synapses within TRN as well as from visual cortex to LIP area. Synaptic strength changes were forced by teaching signal typical for three different age groups (young, middle and elderly) determined experimentally from eye movement data collected by eye tracking device from human subjects preforming a simplified simulated visual navigation task.

STDP Plasticity in TRN within Hierarchical Spike Timing Model of Visual Information Processing

Over the last years, Indoor Localization Systems (ILS) evolved, due to the inability of Global Positioning Systems (GPS) to localize in indoor environments. A variety of studies tackle indoor localization with technologies such as Bluetooth Beacons and RFID that require costly installation, or techniques such as Google Wi-Fi/Cell DB and fingerprinting that leverage from the already existing Wi-FI and telecommunication infrastructure. Additionally, recent studies attempt to solve the same problem using Bio-Inspired techniques, such as Artificial Neural Networks (ANNs) and Deep Neural Networks (DNN). In this paper, we introduce a Multi-Objective Optimization Radiomap Modelling (MOO-RM) based ILS. The MOO-RM ILS divides the dataset into clusters using a K-Means algorithm and trains ANN models on the data of each cluster. The resulting models are fed into a Multi-
Objective Evolutionary Algorithm based on Decomposition (MOEA/D), which minimizes the required storage space and the localization error, simultaneously. Our experimental studies demonstrate the superiority of athe proposed approach on real datasets of Wi-Fi traces with respect to various existing techniques.

Indoor Localization with Multi-Objective Selection of Radiomap Models

Autonomous driving is conducted in complex scenarios, which requires to detect 3D objects in real time scenarios as well as accurately track these 3D objects in order to get such information as location, size, trajectory, velocity. MOT (Multi-Object Tracking) performance is heavily dependent on object detection. Once object detection gives false alarms or missing alarms, the multi-object tracking would be automatically influenced. In this paper, we propose a coupling system which combines 3D object detection and multi-object tracking into one framework. We use the tracked objects as a reference in 3D object detection, in order to locate objects, reduce false or missing alarms in a single frame, and weaken the impact of false and missing alarms on the tracking quality. Our method is evaluated on KITTI dataset and is proved effective.

Robust 3D Detection in Traffic Scenario with Tracking-based Coupling System

In this paper, we present a novel method for jointly detecting and segmenting multiple objects from an untrimmed video. Unlike most existing video object segmentation methods that can only handle a trimmed video in which all video frames contain the target objects, we address a more practical and dicult problem, i.e., joint multi-object detection and segmentation from an untrimmed video where the target objects do not always appear per frame. In particular, our method consists of two modules, i.e., object decision module and object segmentation module. The object decision module is used to detect the objects and decide which target objects need to be separated out from video. As there are usually two or more target objects and they do not always appear in the whole video, we introduce the data association into object decision module to identify their correspondences among frames. The object segmentation module aims to separate the target objects identied by object decision module. In order to extensively evaluate the proposed method, we introduce a new dataset named UNVOSeg dataset, in which 7:2% of the video frames do not contain objects. Experimental results on four datasets demonstrate that our method outperforms most of the state-of-the-art approaches.

Joint Multi-Object Detection and Segmentation from an Untrimmed Video

Large ontologies are available as linked data, and they are used across many domains, but to process them considerable resources are required. RDF provides automation possibilities for semantic interpretation, which can lower the effort. We address the usage of RDF reasoning in large ontologies, and we test approaches for solving reasoning problems, having in mind use cases of low availability of computational resources. In our experiment, we designed and evaluated a method based on a reasoning problem of inferring Schema.org statements from cultural objects described in Wikidata. The method defines two intermediate tasks that reduce the volume of data used during the execution of the RDF reasoner, resulting in an efficient execution taking on average 10.3±7.6 milliseconds per RDF resource. The inferences obtained in the Wikidata test were analysed and found to be correct, and the computational resource requirements for reasoning were significantly reduced. Schema.org inference resulted in at least one rdf:type statement for each cultural resource, but the inference of Schema.org predicates was below expectations. Our experiment on cultural data has shown that Wikidata contains alignment statements to other ontologies used in the cultural domain, which with the application of RDF and OWL reasoning can be used to infer views of Wikidata expressed in cultural domain’s data models.

RDF reasoning on large ontologies: a study on cultural heritage and Wikidata

Brief Presentation of the problem dealt within eTracer. How we formed the knowledge and managed it along with relationships among concepts and how we applied spatio-temporal ruled-based reasoning on top. The presentation also involves examples of implicit relationships constructs, validations examples via SHACL, system-wise evaluation, a simulation example and concludes with future work and references.

Knowledge-base Management and Reasoning on Cultural and Natural Touristic Routes

Modelling of knowledge and actions in AI has advanced over the years but it is still a challenging topic due to the infamous frame problem, the inadequate formalization and the lack of automation. Some problems in cyber security such as logical vulnerability, risk assessment, policy validation etc. still require formal approach. In this presentation we present the foundations of a new formal framework to address these challenges. Our approach is based on three-level formalisation: ontological, logical and analytical levels. Here we are presenting the first two levels which allow to model the security policies and provide a practical solution to the frame problem by efficient utilization of parameters as side effects. Key concepts are the situations, actions, events and rules. Our framework has potential use for analysis of a wide range of transactional systems within the financial, commercial and business domains and further work will include analytical level where we can perform vulnerability analysis of the model.

Ontological Foundations of Modelling Security Policies for Logical Analytics

This work proposes an efficient Content-Based (CB) product recommendation
methodology that promotes diversity. A heuristic CB approach incor- AQ1
porating feature weighting and Locality-Sensitive Hashing (LSH) is used, along
with the TF-IDF method and functionality of tuning the importance of product
features to adjust its logic to the needs of various e-commerce sites. The problem
of efficiently producing recommendations, without compromising similarity,
is addressed by approximating product similarities via the LSH technique. The
methodology is evaluated on two sets with real e-commerce data. The evaluation
of the proposed methodology shows that the produced recommendations can
help customers to continue browsing a site by providing them with the necessary
“next step”. Finally, it is demonstrated that the methodology incorporates recommendation diversity which can be adjusted by tuning the appropriate feature
weights.

Tensor-based CUDA Optimization for ANN Inferencing using Parallel Acceleration on Embedded GPU

Downloads

Next from AIAI 2020

Cross-domain Authorship Attribution Using Pre-trained Language Models

Similar lecture

Indoor Localization with Multi-Objective Selection of Radiomap Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Tensor-based CUDA Optimization for ANN Inferencing using Parallel Acceleration on Embedded GPU

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AIAI 2020

Cross-domain Authorship Attribution Using Pre-trained Language Models

Similar lecture

Indoor Localization with Multi-Objective Selection of Radiomap Models

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads