Model Zoo

DX Model Zoo

Image Classification

Name	Dataset	Input Resolution	Operations	Parameters	License	Metric	Raw Accuracy	NPU Accuracy	FPS	FPS/Watt
AlexNet-1	ImageNet	224x224x3	0.72	61.1	BSD 3-Clause	Top1	56.54	56.44	663	1,218.68
DenseNet121-1	ImageNet	224x224x3	3.19	8.04	BSD 3-Clause	Top1	74.43	73.29	33	0.38
DenseNet161-1	ImageNet	224x224x3	8.43	28.86	MIT	Top1	77.11	76.89	14	0.38
EfficientNetB2-1	ImageNet	288x288x3	1.6	9.08	BSD 3-Clause	Top1	80.61	79.06	757	740.28
EfficientNetV2S-1	ImageNet	384X384x3	9.47	21.38	BSD 3-Clause	Top1	84.24	0.29	624	270.94
HarDNet39DS-1	ImageNet	224x224x3	4.26	17.56	MIT License	Top1	72.08	71.62	1,838	2,204.90
MobileNetV1-1	ImageNet	224x224x3	0.58	4.22	BSD 3-Clause	Top1	69.49	68.84	4,940	2,763.11
MobileNetV2-1	ImageNet	224x224x3	0.32	3.49	MIT	Top1	72.14	71.7	3,617	3,161.57
MobileNetV3L-1	ImageNet	224x224x3	0.23	5.47	BSD 3-Clause	Top1	75.25	72.46	3,258	3,506.84
RegNetX400MF-1	ImageNet	224x224x3	0.42	5.48	BSD 3-Clause	Top1	74.88	74.11	1,500	2,042.85
RegNetX800MF-1	ImageNet	224x224x3	0.81	7.24	BSD 3-Clause	Top1	77.52	77.17	1,087	1,241.74
RegNetY200MF-1	ImageNet	224x224x3	0.21	3.15	BSD 3-Clause	Top1	70.36	69.02	2,577	3,880.45
RegNetY400MF-1	ImageNet	224x224x3	0.41	4.33	MIT	Top1	75.78	74.96	1,781	2,109.10
RegNetY800MF-1	ImageNet	224x224x3	0.85	6.42	BSD 3-Clause	Top1	78.83	77.01	1,217	1,235.36
RepVGGA1-1	ImageNet	320X320x3	4.83	12.79	BSD 3-Clause	Top1	75.28	63.59	1,616	545.90
ResNet101-1	ImageNet	224x224x3	7.8	44.55	BSD 3-Clause	Top1	81.9	81.22	669	265.66
ResNet18-1	ImageNet	224x224x3	1.82	11.69	BSD 3-Clause	Top1	69.75	69.62	2,780	1,114.47
ResNet34-1	ImageNet	224x224x3	3.67	21.79	BSD 3-Clause	Top1	73.3	73.28	1,498	556.93
ResNet50-1	ImageNet	224x224x3	4.27	25.0	BSD 3-Clause	Top1	80.85	80.3	1,138	480.58
ResNeXt26_32x4d-1	ImageNet	224x224x3	2.49	15.37	BSD 3-Clause	Top1	75.85	75.61	842	573.06
ResNeXt50_32x4d-1	ImageNet	224x224x3	4.27	25.0	BSD 3-Clause	Top1	81.19	80.29	490	336.72
SqueezeNet1_0-1	ImageNet	224x224x3	0.83	1.25	AGPL-3.0	Top1	58.09	56.59	2,133	1,597.78
SqueezeNet1_1-1	ImageNet	224x224x3	0.36	1.24	AGPL-3.0	Top1	58.18	56.07	4,177	4,045.80
VGG11-1	ImageNet	224x224x3	7.63	132.86	GPL-3.0	Top1	69.03	68.98	294	238.05
VGG11BN-1	ImageNet	224x224x3	7.63	132.86	GPL-3.0	Top1	70.37	70.22	291	231.96
VGG13-1	ImageNet	224x224x3	11.34	133.05	GPL-3.0	Top1	69.93	69.63	271	188.84
VGG13BN-1	ImageNet	224x224x3	11.34	133.05	AGPL-3.0	Top1	71.55	71.39	272	169.69
VGG19BN-1	ImageNet	224x224x3	19.69	143.67	AGPL-3.0	Top1	74.24	72.98	234	131.04
WideResNet101-2	ImageNet	224x224x3	22.81	126.82	AGPL-3.0	Top1	82.52	82.31	276	96.73
WideResNet50-2	ImageNet	224x224x3	11.43	68.85	AGPL-3.0	Top1	81.61	81.08	485	187.19

Object Detection

Name	Dataset	Input Resolution	Operations	Parameters	License	Metric	Raw Accuracy	NPU Accuracy	FPS	FPS/Watt
SSDMV1-1	VOC2007Detection	300X300x3	1.55	9.48	AGPL-3.0	mAP50	67.59	67.699	1,847	1,256.19
SSDMV2Lite-1	VOC2007Detection	300x300x3	0.7	3.38	AGPL-3.0	mAP50	68.704	68.652	1,641	1,567.31
YOLOV3-1	COCO	640x640x3	81.13	62.02	Apache-2.0	mAP50	46.654	65.719	104	25.70
YOLOV5L-1	COCO	640x640x3	57.1	46.64	MIT	mAP50	48.74	67.146	140	35.95
YOLOV5M-1	COCO	640x640x3	26.07	21.27	MIT	mAP50	45.082	63.895	181	73.83
YOLOV5N-1	COCO	640x640x3	2.71	1.97	GPL-3.0	mAP50	28.081	44.945	304	484.52
YOLOV5S-1	COCO	640x640x3	9.1	7.33	GPL-3.0	mAP50	37.451	56.799	278	201.90
YOLOV7-2	COCO	640x640x3	55.28	36.92	GPL-3.0	mAP50	50.862	69.587	113	37.47
YOLOV7E6-1	COCO	1280x1280x3	269.21	97.27	GPL-3.0	mAP50	55.58	73.221	21	7.36
YOLOV7Tiny-1	COCO	640x640x3	7.01	6.24	GPL-3.0	mAP50	37.289	55.058	269	225.77
YOLOV8L-1	COCO	640x640x3	85.13	43.69	AGPL-3.0	mAP50	52.572	68.567	82	24.17
YOLOv9-C-2	COCO	640x640x3	53.92	25.31	GPL-3.0	mAP50	52.856	68.868	80	35.72
YOLOv9-S-2	COCO	640x640x3	14.51	7.13	GPL-3.0	mAP50	46.683	59.338	167	113.54
YOLOv9-T-2	COCO	640x640x3	4.56	2.03	GPL-3.0	mAP50	38.268	49.704	188	235.46
YOLOXS-1	COCO	640x640x3	14.41	8.96	Apache-2.0	mAP50	40.29	59.042	289	129.56

Segmentation

Name	Dataset	Input Resolution	Operations	Parameters	License	Metric	Raw Accuracy	NPU Accuracy	FPS	FPS/Watt
BiSeNetV1-1	CitySpace	1280x2048x3	118.98	13.27	BSD 3-Clause	mIoU	75.367	75.187	18	14.57
BiSeNetV2-1	CitySpace	1280x2048x3	99.14	3.35	MIT	mIoU	74.925	74.109	31	19.69
DeepLabV3PlusMobilenet-1	VOCSegmentation	512x512x3	26.62	5.8	BSD 3-Clause	mIoU	68.473	50.448	224	73.94

Face ID

Name	Dataset	Input Resolution	Operations	Parameters	License	Metric	Raw Accuracy	NPU Accuracy	FPS	FPS/Watt
YOLOV5M_Face-1	WiderFace	640x640x3	25.84	22.0	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	95.507 94.027 85.649	95.48 94.035 85.765	196	70.76
YOLOV5S_Face-1	WiderFace	640x640x3	8.53	8.07	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	94.57 92.94 83.698	94.616 92.981 83.78	356	213.96
YOLOV7_Face-1	WiderFace	640x640x3	54.63	38.55	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	96.925 95.689 88.337	97.002 95.738 88.323	117	37.55
YOLOV7_W6_Face-1	WiderFace	960x960x3	100.21	74.44	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	96.41 95.091 88.61	96.387 95.085 88.535	69	20.29
YOLOV7_W6_TTA_Face-1	WiderFace	1280x1280x3	178.16	77.96	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	95.89 94.929 89.952	95.977 94.995 90.155	38	11.26
YOLOV7s_Face-1	WiderFace	640x640x3	9.35	6.26	GPL-3.0	AP(Easy) AP(Med) AP(Hard)	94.86 93.3 85.304	94.887 93.266 85.198	263	174.37

Image De-noising

Name	Dataset	Input Resolution	Operations	Parameters	License	Metric	Raw Accuracy	NPU Accuracy	FPS	FPS/Watt
DnCNN-2	BSD68	512x512x1	145.8	0.56	BSD 3-Clause	PSNR SSIM	31.709 0.8905	30.5031 0.8688	37	12.19
DnCNN-3	BSD68	512x512x1	145.8	0.56	MIT	PSNR SSIM	29.1919 0.8276	28.6563 0.816	37	11.98
DnCNN-4	BSD68	512x512x1	145.8	0.56	Apache-2.0	PSNR SSIM	26.1882 0.7184	21.1825 0.6026	37	11.98