Mixed-precision quantization for neural networks based on error limit (<i>Invited</i>)

Fig. 1. (a) Photograph of deep learning convolutional 8-bit quantization procession^[6]; (b) The distribution trend of the most valued weights in the first 20 layers of the YOLOV5 s network; (c) Distribution of activation maximum and cutoff value during network quantization in YOLOV5 s

Download full size

View in Article

Fig. 2. Framework of network hierarchical policy methodology

Download full size

View in Article

Fig. 3. Example of COCO dataset detection results

Download full size

View in Article

Table 1. Product quantization method and shift quantization method
View table
View in Article
Table 1. Product quantization method and shift quantization method
Quantitative method Operation
${q}\left(w,{b}_{i}\right)=round\left(w/s\right)$ Multiplication
${q}\left(w,{b}_{i}\right)=round\left(w×{2}^{fl}\right)$ Displacement

Table 2. The performance of different quantification methods on the VOC2007 dataset
View table
View in Article
Table 2. The performance of different quantification methods on the VOC2007 dataset
Network model Dataset bit mAP.5-.95
Displacement Multiplication
YOLOV5 s VOC 8 63.4% 77.9%
7 26.5% 68.8%
6 4.6% 39.5%
32 81.8%

Table 3. Network accuracy before and after quantization with different truncation methods
View table
View in Article
Table 3. Network accuracy before and after quantization with different truncation methods
bit 8 7 6 5 32
mAP MAX 78.9% 67.4% 46.7% 4.0% 82.6%
MSE 82.7% 76.0% 69.0% 31.7%

Table 4. Error limit parameter γ value comparison
View table
View in Article
Table 4. Error limit parameter γ value comparison
γ Compression radio Average bit mAP
0.08 4.93 6.49 79.6%
0.10 5.13 6.23 77.8%
0.125 5.74 5.57 72.3%
0.142 6.11 5.23 62.8%
0.166 6.31 5.07 63.3%
0.20 7.14 4.48 21.0%

Table 5. Test results of different quantification methods on COCO dataset and VOC2011 dataset

View table

View in Article

Table 5. Test results of different quantification methods on COCO dataset and VOC2011 dataset

Dataset	Method	bit	γ	mAP@0.5	mAP@0.5-0.95	Model size
COCO	Unified bit	7		0.567	0.345	6.35
		6		0.503	0.301	5.45
		5		0.386	0.215	4.54
	Mixed bit	6.49	0.08	0.602	0.368	5.89
		5.57	0.125	0.546	0.322	5.05
		5.07	0.166	0.446	0.260	4.60
	Ori model	32		0.636	0.411	29.07
VOC2011	Unified bit	7		0.950	0.732	6.35
		6		0.925	0.643	5.45
		5		0.533	0.295	4.54
	Mixed bit	6.49	0.08	0.950	0.706	5.89
		5.57	0.125	0.981	0.669	5.05
		5.07	0.166	0.782	0.456	4.60
	Ori model	32		0.950	0.786	29.07

Table 6. VOC2011 dataset category accuracy detection table

View table

View in Article

Table 6. VOC2011 dataset category accuracy detection table

Dataset	Method	bit	mAP@0.5	Aeroplane	Bicycle	Bird	Boat	Bottle	Chair	Dog	Person	Sheep	Train	Tvmonitor
VOC2011	Unite	5	0.782	0.753	0.435	0.497	0.995	0.801	0.995	0.249	0.897	0.995	0.995	0.995
VOC2011	Mixed	5	0.533	0.232	0.324	0.497	0.484	0.209	0.995	0.332	0.455	0.995	0.995	0.34

Tools

Get Citation

Copy Citation Text

Yiduo Li, Zibo Guo, Kai Liu, Xiaoyao Sun. Mixed-precision quantization for neural networks based on error limit (Invited)[J]. Infrared and Laser Engineering, 2022, 51(4): 20220166

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Special issue—Infrared detection and recognition technology under superspeed flow field

Received: Mar. 10, 2022

Accepted: Apr. 11, 2022

Published Online: May. 18, 2022

The Author Email:

DOI:10.3788/IRLA20220166

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Product quantization method and shift quantization method

Table 1. Product quantization method and shift quantization method

Table 2. The performance of different quantification methods on the VOC2007 dataset

Table 2. The performance of different quantification methods on the VOC2007 dataset

Table 3. Network accuracy before and after quantization with different truncation methods

Table 3. Network accuracy before and after quantization with different truncation methods

Table 4. Error limit parameter γ value comparison

Table 4. Error limit parameter γ value comparison

Table 5. Test results of different quantification methods on COCO dataset and VOC2011 dataset

Table 5. Test results of different quantification methods on COCO dataset and VOC2011 dataset

Table 6. VOC2011 dataset category accuracy detection table

Table 6. VOC2011 dataset category accuracy detection table

微信扫一扫：分享