Search

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.

Convolutional Neural Networks (CNNs) are popular models that are widely used in image classification, target recognition, and other fields. Model compression is a common step in transplanting neural networks into embedded devices, and it is often used in the retraining stage. However, it requires a high expenditure of time by retraining weight data to atone for the loss of precision. Unlike in prior designs, we propose a novel model compression approach based on Simonk-means, which is specifically designed to support a hardware acceleration scheme. First, we propose an extension algorithm named Simonk-means based on simple k-means. We use Simonk-means to cluster trained weights in convolutional layers and fully connected layers. Second, we reduce the consumption of hardware resources in data movement and storage by using a data storage and index approach. Finally, we provide the hardware implementation of the compressed CNN accelerator. Our evaluations on several classifications show that our design can achieve 5.27× compression and reduce 74.3% of the multiply–accumulate (MAC) operations in AlexNet on the FASHION-MNIST dataset.

Spiking neural networks (SNN) increasingly attract attention for their similarity to the biological neural system. Hardware implementation of spiking neural networks, however, remains a great challenge due to their excessive complexity and circuit size. This work introduces a novel optimization method for hardware friendly SNN architecture based on a modified rate coding scheme called Binary Streamed Rate Coding (BSRC). BSRC combines the features of both rate and temporal coding. In addition, by employing a built-in randomizer, the BSRC SNN model provides a higher accuracy and faster training. We also present SNN optimization methods including structure optimization and weight quantization. Extensive evaluations with MNIST SNNs demonstrate that the structure optimization of SNN (81-30-20-10) provides 183.19 times reduction in hardware compared with SNN (784-800-10), while providing an accuracy of 95.25%, a small loss compared with 98.89% and 98.93% reported in the previous works. Our weight quantization reduces 32-bit weights to 4-bit integers leading to further hardware reduction of 4 times with only 0.56% accuracy loss. Overall, the SNN model (81-30-20-10) optimized by our method shrinks the SNN&rsquo, s circuit area from 3089.49 mm2 for SNN (784-800-10) to 4.04 mm2&mdash, a reduction of 765 times.