Optimization of Object Detection CNN With Weight Quantization and Scale Factor Consolidation

Al-Hamid, Ali A.; Kim, Taehyeon; Park, TaeMi; Kim, HyungWon

doi:10.1109/icce-asia53811.2021.9641913

Cited by 2 publications

(2 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…YOLOv2-tiny has been adopted in some studies to develop real-time FMD systems. For example, in [157], a weight quantization scheme is presented to design a compact CNN model that detects individuals with or without masks based on YOLOv2-tiny.…”

Section: Lightweight Object Detectorsmentioning

confidence: 99%

Face Mask Detection in Smart Cities Using Deep and Transfer Learning: Lessons Learned from the COVID-19 Pandemic

Himeur

Al-Maadeed

Varlamis

et al. 2023

Systems

View full text Add to dashboard Cite

After different consecutive waves, the pandemic phase of Coronavirus disease 2019 does not look to be ending soon for most countries across the world. To slow the spread of the COVID-19 virus, several measures have been adopted since the start of the outbreak, including wearing face masks and maintaining social distancing. Ensuring safety in public areas of smart cities requires modern technologies, such as deep learning and deep transfer learning, and computer vision for automatic face mask detection and accurate control of whether people wear masks correctly. This paper reviews the progress in face mask detection research, emphasizing deep learning and deep transfer learning techniques. Existing face mask detection datasets are first described and discussed before presenting recent advances to all the related processing stages using a well-defined taxonomy, the nature of object detectors and Convolutional Neural Network architectures employed and their complexity, and the different deep learning techniques that have been applied so far. Moving on, benchmarking results are summarized, and discussions regarding the limitations of datasets and methodologies are provided. Last but not least, future research directions are discussed in detail.

show abstract

Section: Lightweight Object Detectorsmentioning

confidence: 99%

Face Mask Detection in Smart Cities Using Deep and Transfer Learning: Lessons Learned from the COVID-19 Pandemic

Himeur

Al-Maadeed

Varlamis

et al. 2023

Systems

View full text Add to dashboard Cite

show abstract

“…As they focus on models that utilize activation functions that are mere clamps (e.g., ReLU and ReLU6), our proposed method can improve the quantized models' accuracy without requiring further retraining. In [33], weight quantization and scale factor consolidation were evaluated using a modified YOLOv2-Tiny model with the mask and no-mask datasets. The evaluation demonstrated that it saves more than 50% of the parameter memory and 56.21% of the inference computation using 16-bit for quantization, which is not efficient for some hardware-constrained devices.…”

Section: Related Workmentioning

confidence: 99%

Unified Scaling-Based Pure-Integer Quantization for Low-Power Accelerator of Complex CNNs

Al-Hamid

Kim

2023

Electronics

Self Cite

View full text Add to dashboard Cite

Although optimizing deep neural networks is becoming crucial for deploying the networks on edge AI devices, it faces increasing challenges due to scarce hardware resources in modern IoT and mobile devices. This study proposes a quantization method that can quantize all internal computations and parameters in the memory modification. Unlike most previous methods that primarily focused on relatively simple CNN models for image classification, the proposed method, Unified Scaling-Based Pure-Integer Quantization (USPIQ), can handle more complex CNN models for object detection. USPIQ aims to provide a systematic approach to convert all floating-point operations to pure-integer operations in every model layer. It can significantly reduce the computational overhead and make it more suitable for low-power neural network accelerator hardware consisting of pure-integer datapaths and small memory aimed at low-power consumption and small chip size. The proposed method optimally calibrates the scale parameters for each layer using a subset of unlabeled representative images. Furthermore, we introduce a notion of the Unified Scale Factor (USF), which combines the conventional two-step scaling processes (quantization and dequantization) into a single process for each layer. As a result, it improves the inference speed and the accuracy of the resulting quantized model. Our experiment on YOLOv5 models demonstrates that USPIQ can significantly reduce the on-chip memory for parameters and activation data by ~75% and 43.68%, respectively, compared with the floating-point model. These reductions have been achieved with a minimal loss in mAP@0.5—at most 0.61%. In addition, our proposed USPIQ exhibits a significant improvement in the inference speed compared to ONNX Run-Time quantization, achieving a speedup of 1.64 to 2.84 times. We also demonstrate that USPIQ outperforms the previous methods in terms of accuracy and hardware reduction for 8-bit quantization of all YOLOv5 versions.

show abstract

Optimization of Object Detection CNN With Weight Quantization and Scale Factor Consolidation

Cited by 2 publications

References 3 publications

Face Mask Detection in Smart Cities Using Deep and Transfer Learning: Lessons Learned from the COVID-19 Pandemic

Face Mask Detection in Smart Cities Using Deep and Transfer Learning: Lessons Learned from the COVID-19 Pandemic

Unified Scaling-Based Pure-Integer Quantization for Low-Power Accelerator of Complex CNNs

Contact Info

Product

Resources

About