3.2 KiB

Raw Blame History

Inception系列

概述

正在持续更新中...... GoogLeNet在预测时，图像的crop_size设置为224，resize_short_size设置为256，其余模型在预测时，图像的crop_size设置为299，resize_short_size设置为320。

精度、FLOPS和参数量

Models	Top1	Top5	Reference top1	Reference top5	FLOPS (G)	Parameters (M)
GoogLeNet	0.707	0.897	0.698		2.880	8.460
Xception41	0.793	0.945	0.790	0.945	16.740	22.690
Xception41 _deeplab	0.796	0.944			18.160	26.730
Xception65	0.810	0.955			25.950	35.480
Xception65 _deeplab	0.803	0.945			27.370	39.520
Xception71	0.811	0.955			31.770	37.280
InceptionV4	0.808	0.953	0.800	0.950	24.570	42.680

FP16预测速度

Models	batch_size=1 (ms)	batch_size=4 (ms)	batch_size=8 (ms)	batch_size=32 (ms)
GoogLeNet	1.428	1.833	2.138	4.143
Xception41	1.545	2.772	4.961	18.447
Xception41 _deeplab	1.630	2.647	4.462	16.354
Xception65	5.398	4.215	8.611	28.702
Xception65 _deeplab	5.317	3.688	6.168	23.108
Xception71	2.732	5.033	8.948	33.857
InceptionV4	6.172	7.558	9.527	24.021

FP32预测速度

Models	batch_size=1 (ms)	batch_size=4 (ms)	batch_size=8 (ms)	batch_size=32 (ms)
GoogLeNet	1.436	2.904	3.800	9.049
Xception41	3.402	7.889	14.953	56.142
Xception41 _deeplab	3.778	8.396	15.449	58.735
Xception65	6.802	13.935	34.301	87.256
Xception65 _deeplab	8.583	12.132	22.917	87.983
Xception71	6.156	14.359	27.360	107.282
InceptionV4	10.384	17.438	23.312	68.777