Error Modeling for Image Processing Filters accelerated onto SRAM-based FPGAs

IEEE Trans. Emerg. Topics Comput.

et al. 2021

Self Cite

Image processing applications exhibit an intrinsic degree of fault tolerance due to i) the redundant nature of images, and ii) the possible ability of the consumers of the application output to effectively carry out their task even when it is slightly corrupted. In this application scenario the classical Duplication with Comparison (DWC) scheme, that rejects images (and requires re-executions) when the two replicas' outputs differ in a per-pixel comparison, may be over-conservative. In this paper, we propose a novel lightweight fault tolerant scheme specifically tailored for image processing applications. The proposed scheme enhances the state-of-the-art by: i) improving the DWC scheme by replacing one of the two exact replicas with an approximated counterpart, and ii) allowing to distinguish between usable and unusable images instead of corrupted and uncorrupted ones by means of a Convolutional Neural Network-based checker.To tune the proposed scheme we introduce a specific design methodology that optimizes both execution time and fault detection capability of the hardened system. We report the results of the application of the proposed approach on two case studies; our proposal achieves an average execution time reduction larger than 30% w.r.t. the DWC with re-execution, and less than 4% misclassified unusable images.

Section: Methodsmentioning

confidence: 99%

Approximation-based Fault Tolerance in Image Processing Applications

Biasielli¹,

IEEE Trans. Emerg. Topics Comput.

et al. 2021

Self Cite

“…We here adopt a strategy similar to the one proposed in [2], [9] to generate a large set of corrupted images corresponding to the adopted fault model (e.g. SEUs), target platform and image processing application being executed.…”

Section: Fault Corruption Analysismentioning

confidence: 99%

Fault Impact Estimation for Lightweight Fault Detection in Image Filtering

Boracchi

et al. 2022

IEEE Trans. Comput.

Self Cite

Classical redundancy-based fault detection techniques, such as Duplication with Comparison (DWC), rely on replicating the computation and comparing the replicas' output at a bit-wise granularity. In many application environments these costs are prohibitive, especially when applications are characterized by an intrinsic level of tolerance. This paper presents a novel fault-detection approach for the specific context of image filtering. Peculiarity of the proposed approach is that it estimates the impact of the fault on the processed output, in order to determine whether the image is usable or should be re-processed. To limit overheads, the proposed solution exploits Approximate Computing (AC), allowing the definition of disciplined AC strategies to trade-off between accuracy and costs. Core of our solution is the successful combination of Image Quality Assessment metrics and Machine Learning models to assess the visual impact of the fault in a lightweight manner. Extensive experimental campaigns demonstrate the effectiveness of the solution, achieving achieving a reduction in terms of execution time up to 44% with respect to the classical DWC, with a fault detection precision ranging from 94.58% to 96.70%, and recall ranging from 88.2% to 97.8%, depending on the adopted level of approximation.

“…It receives from the host machine, through the serial connection, the input image to be used, the golden output and the mask presenting all injectable memory locations and performs a fault injection campaign by following a classical execution flow, where all injectable memory locations are corrupted one at a time, and returns i) the response for each experiment (corrupted output, not corrupted output ,or timeout), and ii) the corrupted output images to the host machine. As a final note, as already discussed in [11], numerous input images are employed during the fault injection campaign not to introduce biases in the identified error models.…”

Section: A Circuit-level Fault Injectionmentioning

confidence: 99%

“…It is indeed fundamental that the corrupted data injected in the application well represents the effects of the real possible faults, i.e., that the adopted error models are accurate. A first attempt of extracting accurate error models specific for image processing applications accelerated onto SRAM-based FPGAs has been presented in [11].…”

Section: Introduction and Related Workmentioning

confidence: 99%

“…• application-level error simulation, and • usability-based reliability analysis. The first concept is achieved by means of a two-step process based on i) preliminary fault injection campaigns used to define error model libraries for the most common image processing filters based on the approach presented in [11], and ii) error simulation of the entire image processing application where the previously identified error models are exploited. The second concept requires the definition of a scenario-based oracle capable of modeling the main characteristics of the downstream control application and to classify whether the (possibly corrupted) output of the image processing application would still allow the downstream control application to take correct decisions, or not.…”

Section: Introduction and Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Usability-based Cross-Layer Reliability Evaluation of Image Processing Applications

2021 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT)

Mazzeo

et al. 2021

Self Cite

Image processing applications are today increasingly employed in safety-and mission-critical fields for perception tasks. It is therefore vital to analyse the reliability of the designed system before its deployment and, if necessary, to adopt specific hardening techniques. In this paper we propose a cross-layer reliability evaluation framework specifically meant for image processing applications accelerated onto SRAM-based FPGAs. The framework is based on two key concepts: i) an applicationlevel error simulation based on validated error models to speedup execution times, and ii) an analysis of the usability of the output images based on the working scenario. Such usability analysis allows the designer to study whether the downstream system would be able to take correct decisions even if the image processing outputs are corrupted. We applied the proposed idea on a motion detection application and we compared the achieved accuracy and the required execution times with the ones of a circuit-level fault injector, here considered as a ground truth. This experiment highlighted an accuracy comparable with the one of the fault injection with a dramatic time saving.