A Novel Memory-Scheduling Strategy for Large Convolutional Neural Network on Memory-Limited Devices

Table 6

Memory usage in each layer for VGG-D19.

AlexNet layer	Memory limit(MB)
	200	100	50	10	5
1	0.58	0.58	0.58	0.58	0.58
2	12.82	12.82	12.82	6.41	4.27
3	12.26	12.26	12.26	6.13	4.09
4	24.50	24.50	24.50	8.17	4.90
5	12.53	12.53	12.53	6.41	4.37
6	15.31	15.31	15.31	7.66	3.83
7	9.19	9.19	9.19	9.19	4.59
8	6.69	6.69	6.69	6.69	3.63
9	12.25	12.25	12.25	6.13	4.08
10	7.25	7.25	7.25	7.25	3.44
11	7.66	7.66	7.66	7.66	3.83
12	4.59	4.59	4.59	4.59	4.59
13	5.31	5.31	5.31	5.31	4.19
14	6.13	6.13	6.13	6.13	3.06
15	5.31	5.31	5.31	5.31	4.19
16	6.13	6.13	6.13	6.13	3.06
17	5.31	5.31	5.31	5.31	4.19
18	6.13	6.13	6.13	6.13	3.06
19	7.56	7.56	7.56	7.56	4.19
20	3.83	3.83	3.83	3.83	3.83
21	2.30	2.30	2.30	2.30	2.30
22	10.53	10.53	10.53	6.03	3.33
23	3.06	3.06	3.06	3.06	3.06
24	10.53	10.53	10.53	6.03	3.33
25	3.06	3.06	3.06	3.06	3.06
26	10.53	10.53	10.53	6.03	3.33
27	3.06	3.06	3.06	3.06	3.06
28	10.53	10.53	10.53	6.03	4.53
29	1.91	1.91	1.91	1.91	1.91
30	0.77	0.77	0.77	0.77	0.77
31	9.38	9.38	9.38	9.38	3.38
32	0.77	0.77	0.77	0.77	0.77
33	9.38	9.38	9.38	9.38	3.38
34	0.77	0.77	0.77	0.77	0.77
35	9.38	9.38	9.38	9.38	3.38
36	0.77	0.77	0.77	0.77	0.77
37	196.39	98.39	0.38	0.38	0.38
38	0.48	0.48	0.48	0.48	0.48
39	0.11	0.11	49.11	9.91	4.95
40	64.03	64.03	32.02	0.02	0.02
41	0.03	0.03	0.03	9.18	4.96
42	15.64	15.64	15.64	7.83	3.92
43	0.02	0.02	0.02	0.02	0.02