Browse ORBilu by Open Access
- What it is and what it isn't
- Green Road / Gold Road?
- Ready to Publish. Now What?
- How can I support the OA movement?
- Where can I learn more?
ORBilu is a project developed by
   Profiling the real world potential of neural network compressionLorentz, Joe  ; ; et alin 2022 IEEE International Conference on Omni-layer Intelligent Systems (COINS), Barcelona 1-3 August 2022 (2022, August 01) Abstract—Many real world computer vision applications are required to run on hardware with limited computing power, often referred to as ”edge devices”. The state of the art in computer vision continues ... [more ▼] Abstract—Many real world computer vision applications are required to run on hardware with limited computing power, often referred to as ”edge devices”. The state of the art in computer vision continues towards ever bigger and deeper neural networks with equally rising computational requirements. Model compression methods promise to substantially reduce the computation time and memory demands with little to no impact on the model robustness. However, evaluation of the compression is mostly based on theoretic speedups in terms of required floating-point operations. This work offers a tool to profile the actual speedup offered by several compression algorithms. Our results show a significant discrepancy between the theoretical and actual speedup on various hardware setups. Furthermore, we show the potential of model compressions and highlight the importance of selecting the right compression algorithm for a target task and hardware. The code to reproduce our experiments is available at https://hub.datathings.com/papers/2022-coins. [less ▲] Detailed reference viewed: 73 (7 UL)Explaining Defect Detection with Saliency MapsLorentz, Joe ; ; et alin 34th International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2021, Kuala Lumpur, Malaysia, July 26–29, 2021, Proceedings, Part II (2021, July 19) The rising quality and throughput demands of the manufacturing domain require flexible, accurate and explainable computer-vision solutions for defect detection. Deep Neural Networks (DNNs) reach state-of ... [more ▼] The rising quality and throughput demands of the manufacturing domain require flexible, accurate and explainable computer-vision solutions for defect detection. Deep Neural Networks (DNNs) reach state-of-the-art performance on various computer-vision tasks but wide-spread application in the industrial domain is blocked by the lacking explainability of DNN decisions. A promising, human-readable solution is given by saliency maps, heatmaps highlighting the image areas that influence the classifier’s decision. This work evaluates a selection of saliency methods in the area of industrial quality assurance. To this end we propose the distance pointing game, a new metric to quantify the meaningfulness of saliency maps for defect detection. We provide steps to prepare a publicly available dataset on defective steel plates for the proposed metric. Additionally, the computational complexity is investigated to determine which methods could be integrated on industrial edge devices. Our results show that DeepLift, GradCAM and GradCAM++ outperform the alternatives while the computational cost is feasible for real time applications even on edge devices. This indicates that the respective methods could be used as an additional, autonomous post-classification step to explain decisions taken by intelligent quality assurance systems. [less ▲] Detailed reference viewed: 150 (19 UL)The Next Evolution of MDE: A Seamless Integration of Machine Learning into Domain ModelingHartmann, Thomas ; ; Fouquet, François et alin Software and Systems Modeling (2017) Machine learning algorithms are designed to resolve unknown behaviors by extracting commonalities over massive datasets. Unfortunately, learning such global behaviors can be inaccurate and slow for ... [more ▼] Machine learning algorithms are designed to resolve unknown behaviors by extracting commonalities over massive datasets. Unfortunately, learning such global behaviors can be inaccurate and slow for systems composed of heterogeneous elements, which behave very differently, for instance as it is the case for cyber-physical systems andInternet of Things applications. Instead, to make smart deci-sions, such systems have to continuously refine the behavior on a per-element basis and compose these small learning units together. However, combining and composing learned behaviors from different elements is challenging and requires domain knowledge. Therefore, there is a need to structure and combine the learned behaviors and domain knowledge together in a flexible way. In this paper we propose to weave machine learning into domain modeling. More specifically, we suggest to decompose machine learning into reusable, chainable, and independently computable small learning units, which we refer to as microlearning units.These micro learning units are modeled together with and at the same level as the domain data. We show, based on asmart grid case study, that our approach can be significantly more accurate than learning a global behavior, while the performance is fast enough to be used for live learning. [less ▲] Detailed reference viewed: 412 (13 UL)Industrial defect detection on the edge with deep learning over scarcely labeled and extremely imbalanced dataLorentz, Joe ; ; et alE-print/Working paper (n.d.) Abstract—Reliable automated defect detection is an integral part of modern manufacturing and improved performance can provide a competitive advantage. Despite the proven capabilities of convolutional ... [more ▼] Abstract—Reliable automated defect detection is an integral part of modern manufacturing and improved performance can provide a competitive advantage. Despite the proven capabilities of convolutional neural networks (CNNs) for image classification, application on real world tasks remains challenging due to the high demand for labeled and well balanced data of the common supervised learning scheme. Semi-supervised learning (SSL) promises to achieve comparable accuracy while only requiring a small fraction of the training samples to be labeled. However, SSL methods struggle with data imbalance and existing benchmarks do not reflect the challenges of real world applications. In this work we present a CNN-based defect detection unit for thermal sensors. We describe how to collect data from a running process and release our dataset of 1k labeled and 293k unlabeled samples. Furthermore, we investigate the use of SSL under this challenging real world task. [less ▲] Detailed reference viewed: 213 (2 UL) |
||
