Tiny Machine Learning for High Accuracy Product Quality Inspection

Albanese, Andrea; Nardello, Matteo; Fiacco, Gianluca; Brunelli, Davide

doi:10.1109/jsen.2022.3225227

Cited by 22 publications

(4 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Three different DNNs have been selected and trained with the dataset developed in Section IV-A. We use state-of-theart architectures, namely MobileNetV2 [39] (alpha parameter equal to 0.35), MobileNetV3 small [40] (alpha parameter equal to 0.35), and SqueezeNet [41] as they show an optimal tradeoff between performance and computational complexity [42], [43] for similar mobile and/or constrained deployment contexts. The DNNs are trained on an NVIDIA GeForce RTX 4090 GPU with the following hyperparameters: II summarizes the number of parameters and the memory footprint of the three architectures.…”

Section: B Deep Neural Network Specificationsmentioning

confidence: 99%

Automated Pest Detection with DNN on the Edge for Precision Agriculture

Albanese¹,

Nardello²,

Brunelli³

2021

Preprint

View full text Add to dashboard Cite

Artificial intelligence has smoothly penetrated several economic activities, especially monitoring and control applications, including the agriculture sector. However, research efforts toward low-power sensing devices with fully functional machine learning (ML) on-board are still fragmented and limited in smart farming. Biotic stress is one of the primary causes of crop yield reduction. With the development of deep learning in computer vision technology, autonomous detection of pest infestation through images has become an important research direction for timely crop disease diagnosis. This paper presents an embedded system enhanced with ML functionalities, ensuring continuous detection of pest infestation inside fruit orchards. The embedded solution is based on a low-power embedded sensing system along with a Neural Accelerator able to capture and process images inside common pheromone-based traps. Three different ML algorithms have been trained and deployed, highlighting the capabilities of the platform. Moreover, the proposed approach guarantees an extended battery life thanks to the integration of energy harvesting functionalities. Results show how it is possible to automate the task of pest infestation for unlimited time without the farmer's intervention.

show abstract

Section: B Deep Neural Network Specificationsmentioning

confidence: 99%

Automated Pest Detection with DNN on the Edge for Precision Agriculture

Albanese¹,

Nardello²,

Brunelli³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Albanese et al deployed two compact CNNs in smart cameras for defect detection in plastic parts. In this context, the MobileNetV2 network achieved a 99% classification accuracy (Albanese et al, 2022). Hu et al (2022) enhanced the VGG16 CNN, achieving a 96.67% accuracy in quality inspection.…”

Section: Introductionmentioning

confidence: 99%

In-situ quality inspection system of injection parts based on transfer learning

Yang,

Shan,

Jin

et al. 2024

RIA

View full text Add to dashboard Cite

Purpose This paper aims to realize an in-situ quality inspection system rapidly for new injection molding (IM) tasks via transfer learning (TL) approach and automation technology. Design/methodology/approach The proposed in-situ quality inspection system consists of an injection machine, USB camera, programmable logic controller and personal computer, interconnected via OPC or USB communication interfaces. This configuration enables seamless automation of the IM process, real-time quality inspection and automated decision-making. In addition, a MobileNet-based deep learning (DL) model is proposed for quality inspection of injection parts, fine-tuned using the TL approach. Findings Using the TL approach, the MobileNet-based DL model demonstrates exceptional performance, achieving validation accuracy of 99.1% with the utilization of merely 50 images per category. Its detection speed and accuracy surpass those of DenseNet121-based, VGG16-based, ResNet50-based and Xception-based convolutional neural networks. Further evaluation using a random data set of 120 images, as assessed through the confusion matrix, attests to an accuracy rate of 96.67%. Originality/value The proposed MobileNet-based DL model achieves higher accuracy with less resource consumption using the TL approach. It is integrated with automation technologies to build the in-situ quality inspection system of injection parts, which improves the cost-efficiency by facilitating the acquisition and labeling of task-specific images, enabling automatic defect detection and decision-making online, thus holding profound significance for the IM industry and its pursuit of enhanced quality inspection measures.

show abstract

“…However, it is important to note that while these works have evaluated their model performances on computers, there is a gap in the literature concerning the assessment of deep learning models, including Recurrent Neural Networks (RNN), LSTM, CNN, and CNN-LSTM, on microcontrollers. It is worth mentioning that there are instances in the literature where deep learning models have been successfully implemented on low-power microcontrollers for various applications, such as spoken keyword spotting [37], human activity recognition [38], [39], gesture recognition [40], acoustic event detection [41], post-stroke assistance and rehabilitation [42], flow analysis in public spaces [43], nutrition monitoring [44], sports activities classification [45], and even for industrial applications [46], [47]. The performance of these models, in terms of power consumption, accuracy, precision, and sensitivity, varies based on the specific models and problems they address.…”

Section: Introductionmentioning

confidence: 99%

Analyzing and Comparing Deep Learning Models on an ARM 32 Bits Microcontroller for Pre-Impact Fall Detection

Benoit,

Escriba,

Gauchard

et al. 2024

IEEE Sensors J.

View full text Add to dashboard Cite

Automated pre-impact fall detection in real-time using raw data acquired from wearable sensors, such as tri-axial accelerometers, remains an open research problem in the context of elderly care. This paper presents a comparative study of nine neural network models, including Dense, CNN, LSTM, GRU, BiLSTM, BiGRU, CNN Dense, CNN LSTM, and CNN GRU, for predicting an occurring fall before impact with the ground using accelerometer data. The models were optimized using Keras Tuner with the TensorFlow backend, a dominant deep learning software framework. Machine learning classifiers suitable for execution on microcontrollers were built and evaluated based on performance metrics such as accuracy, sensitivity, specificity, storage size, inference time, and energy consumption. The results highlight that the CNN DENSE algorithm provides the best detection accuracy (94.70%) with a lead time of 76.91 ms. Sensitivity and specificity reach 95.33% and 94.18%. The energy consumption and inference time are 6.72 mA and 12.88 ms, respectively. In conclusion, deep learning demonstrates increased classification accuracy and a simplified software architecture, but it comes at the cost of decreased energy efficiency and inference speed. These factors are crucial considerations in developing on-demand fall injury prevention wearable systems.

show abstract

Tiny Machine Learning for High Accuracy Product Quality Inspection

Cited by 22 publications

References 22 publications

Automated Pest Detection with DNN on the Edge for Precision Agriculture

Automated Pest Detection with DNN on the Edge for Precision Agriculture

In-situ quality inspection system of injection parts based on transfer learning

Analyzing and Comparing Deep Learning Models on an ARM 32 Bits Microcontroller for Pre-Impact Fall Detection

Contact Info

Product

Resources

About