Model Zoo

DX Model Zoo

Quantization Guide

Q-Lite (Standard): Standard INT8 quantization optimized for fast inference. Recommended as the default choice.
Q-Pro (Advanced): High-precision quantization with fine-tuning to maximize accuracy. (Note: Requires longer compilation time.)

Image Classification

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
AlexNet	ImageNet	224x224x3	715.98	61.10	BSD 3-Clause	Top1	56.538		56.33			56.48			635	1,345.87
DenseNet121	ImageNet	224x224x3	3.18	8.04	BSD 3-Clause	Top1	74.434		74.09			-	-	-	43	140.3
DenseNet161	ImageNet	224x224x3	8.43	28.86	BSD 3-Clause	Top1	77.108		76.10			-	-	-	17	60.59
EfficientNetB2	ImageNet	288x288x3	1.60	9.08	BSD 3-Clause	Top1	80.606		79.79			79.89			767	832.19
EfficientNetLite0	ImageNet	224x224x3	404.52	4.63	Apache-2.0	Top1	67.280		66.19			67.26			3,326	2,863.29
EfficientNetLite1	ImageNet	240x240x3	629.42	5.39	Apache-2.0	Top1	70.950		71.18			-	-	-	2,585	1911.034
EfficientNetLite2	ImageNet	260x260x3	896.92	6.06	Apache-2.0	Top1	71.140		71.03			71.11			1,641	1,338.04
EfficientNetLite3	ImageNet	300x300x3	1.67	8.16	Apache-2.0	Top1	75.310		75.15			75.51			1,058	803.56
EfficientNetLite4	ImageNet	380x380x3	4.08	12.95	Apache-2.0	Top1	77.830		77.38			77.5			530	361.05
EfficientNetV2S	ImageNet	384x384x3	9.47	21.38	BSD 3-Clause	Top1	84.238		81.85			82.84			446	214.99
HarDNet39DS	ImageNet	224x224x3	438.76	3.48	MIT	Top1	72.080		71.60			71.68			1,799	2,437.58
HarDNet68	ImageNet	224x224x3	4.26	17.56	MIT	Top1	76.474		76.41			76.39			578	440.22
InceptionV1	ImageNet	224x224x3	1.52	6.62	Apache-2.0	Top1	70.070		69.98			70.09			2,260	1,303.22
MobileNetV1	ImageNet	224x224x3	578.88	4.22	No License	Top1	69.492		68.91			-	-	-	4,784	3,048.47
MobileNetV2	ImageNet	224x224x3	319.95	3.49	BSD 3-Clause	Top1	72.142		71.83			72.06			3,605	3,404.57
MobileNetV3Large	ImageNet	224x224x3	232.57	5.47	BSD 3-Clause	Top1	75.256		72.01			73.94			3,127	3,785.49
OSNet0_25	ImageNet	224x224x3	135.39	713.14	No License	Top1	58.336		54.30			54.88			1,642	2,993.75
OSNet0_5	ImageNet	224x224x3	436.04	1.14	No License	Top1	69.446		61.72			63.93			1,519	1,979.46
RegNetX400MF	ImageNet	224x224x3	420.64	5.48	BSD 3-Clause	Top1	74.884		74.38			74.49			1,373	2,215.5
RegNetX800MF	ImageNet	224x224x3	810.70	7.24	BSD 3-Clause	Top1	77.522		76.97			77.29			1,023	1,323.75
RegNetY200MF	ImageNet	224x224x3	204.91	3.15	MIT	Top1	70.360		69.65			70.05			2,422	4,146.33
RegNetY400MF	ImageNet	224x224x3	411.81	4.33	BSD 3-Clause	Top1	75.782		74.99			75.54			1,653	2,247.6
RegNetY800MF	ImageNet	224x224x3	847.84	6.42	BSD 3-Clause	Top1	78.828		78.30			-	-	-	1,173	1,298.85
RepVGGA1	ImageNet	320x320x3	4.83	12.79	MIT	Top1	74.090		62.93			74.77			1,593	593.65
RepVGGA2	ImageNet	320x320x3	10.45	25.50	MIT	Top1	76.266		57.70			76.35			818	284.07
ResNet101	ImageNet	224x224x3	7.84	44.50	BSD 3-Clause	Top1	81.898		81.61			81.65			639	288.18
ResNet18	ImageNet	224x224x3	1.82	11.68	BSD 3-Clause	Top1	69.754		68.82			69.64			2,377	1,242.67
ResNet34	ImageNet	224x224x3	3.67	21.79	BSD 3-Clause	Top1	73.294		73.24			73.27			1,373	612.75
ResNet50	ImageNet	224x224x3	4.12	25.53	BSD 3-Clause	Top1	80.854		79.47			80.69			1,069	540.67
ResNeXt26_32x4d	ImageNet	224x224x3	2.49	15.37	MIT	Top1	75.852		74.40			75.66			861	647.07
ResNeXt50_32x4d	ImageNet	224x224x3	4.27	24.99	BSD 3-Clause MIT	Top1	81.190		79.94			80.96			500	376.77
SqueezeNet1_0	ImageNet	224x224x3	832.77	1.25	BSD 3-Clause	Top1	58.088		57.09			-	-	-	2,174	1,716.1
SqueezeNet1_1	ImageNet	224x224x3	357.48	1.24	BSD 3-Clause	Top1	58.180		57.25			-	-	-	4,577	4,495.63
VGG11	ImageNet	224x224x3	7.63	132.86	BSD 3-Clause	Top1	69.034		68.80			68.96			287	263.1
VGG11BN	ImageNet	224x224x3	7.63	132.86	BSD 3-Clause	Top1	70.372		70.21			70.27			287	266.39
VGG13	ImageNet	224x224x3	11.34	133.05	BSD 3-Clause	Top1	69.934		69.85			69.89			268	185.56
VGG13BN	ImageNet	224x224x3	11.34	133.05	BSD 3-Clause	Top1	71.556		71.60			-	-	-	267	206.84
VGG19BN	ImageNet	224x224x3	19.67	143.67	BSD 3-Clause	Top1	74.238		74.17			74.22			235	141.56
WideResNet101_2	ImageNet	224x224x3	22.80	126.82	BSD 3-Clause	Top1	82.520		82.40			-	-	-	271	109.57
WideResNet50_2	ImageNet	224x224x3	11.43	68.85	BSD 3-Clause	Top1	81.610		80.85			81.44			494	210.33

Object Detection

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
DamoYOLO	COCO	640x640x3	50.07	42.06	Apache-2.0	mAP	50.839		48.769			49.161			93	856.48
DamoYoloM	COCO	640x640x3	31.84	28.19	Apache-2.0	mAP	48.816		48.234			48.349			118	66.67
DamoYoloS	COCO	640x640x3	18.96	16.27	Apache-2.0	mAP	46.524		45.724			46.014			131	110.88
DamoYoloT	COCO	640x640x3	9.13	8.50	Apache-2.0	mAP	42.543		42.122			42.477			149	183.69
NanoDet	COCO	416x416x3	5.66	6.74	Apache-2.0	mAP	25.031		24.044			24.874			832	425.37
NanoDet_RepVGGA	COCO	640x640x3	21.44	10.79	Apache-2.0	mAP	29.330		29.021			29.187			391	129.01
SSDMV1	VOC2007Detection	300x300x3	1.55	9.46	MIT	mAP50	67.590		67.638			-	-	-	1,796	1,359.04
SSDMV2Lite	VOC2007Detection	300x300x3	700.57	3.36	MIT	mAP50	68.704		68.652			68.744			1,579	1,632.63
YOLO26l	COCO	640x640x3	46.42	24.85	AGPL-3.0	mAP	53.479		52.997			-	-	-	61	43.35
YOLO26m	COCO	640x640x3	36.57	20.45	AGPL-3.0	mAP	51.824		51.206			-	-	-	81	56.1
YOLO26n	COCO	640x640x3	3.35	2.45	AGPL-3.0	mAP	39.932		39.206			-	-	-	163	319.39
YOLO26s	COCO	640x640x3	11.63	9.54	AGPL-3.0	mAP	47.307		46.701			-	-	-	111	139.51
YOLO26x	COCO	640x640x3	101.72	55.77	AGPL-3.0	mAP	55.546		55.163			-	-	-	34	19.76
YoloV10B	COCO	640x640x3	48.87	19.11	AGPL-3.0	mAP	52.111		40.299			51.453			84	44.2
YoloV10L	COCO	640x640x3	63.65	24.41	AGPL-3.0	mAP	52.871		49.749			52.115			72	34.06
YoloV10M	COCO	640x640x3	31.74	15.40	AGPL-3.0	mAP	50.855		48.789			50.059			93	60.98
YoloV10N	COCO	640x640x3	4.02	2.34	AGPL-3.0	mAP	38.373		35.945			37.992			222	328.27
YoloV10S	COCO	640x640x3	12.04	7.29	AGPL-3.0	mAP	46.046		42.786			45.141			124	127.42
YoloV10X	COCO	640x640x3	85.05	29.52	AGPL-3.0	mAP	54.029		51.566			53.399			45	23.26
YoloV11L	COCO	640x640x3	46.61	25.38	AGPL-3.0	mAP	52.320		51.971			52.05			70	44.76
YoloV11M	COCO	640x640x3	36.40	20.13	AGPL-3.0	mAP	50.519		50.142			50.12			100	58.91
YoloV11N	COCO	640x640x3	3.88	2.66	AGPL-3.0	mAP	38.635		38.34			38.447			217	343.82
YoloV11S	COCO	640x640x3	11.90	9.49	AGPL-3.0	mAP	45.847		45.082			45.755			146	151.84
YoloV11X	COCO	640x640x3	102.16	56.96	AGPL-3.0	mAP	53.560		53.178			53.319			39	20.33
YoloV3	COCO	640x640x3	81.13	61.92	AGPL-3.0	mAP	46.654		46.467			46.432			109	29.22
YoloV5L	COCO	640x640x3	57.10	46.53	AGPL-3.0	mAP	48.740		48.485			48.506			139	40.98
YoloV5M	COCO	640x640x3	26.07	21.17	AGPL-3.0	mAP	45.082		44.803			44.821			211	83.52
YoloV5N	COCO	640x640x3	2.71	1.87	AGPL-3.0	mAP	28.081		27.338			27.646			316	565.2
YoloV5S	COCO	640x640x3	9.10	7.23	AGPL-3.0	mAP	37.451		37.111			37.158			284	225.65
YoloV6N	COCO	640x640x3	5.64	4.32	Apache-2.0	mAP	34.722		31.168			33.572			419	374.24
YoloV7	COCO	640x640x3	55.28	36.92	GPL-3.0	mAP	50.860		50.884			50.823			117	42.27
YoloV7E6	COCO	1280x1280x3	269.21	97.20	GPL-3.0	mAP	55.216		55.617			55.472			21	8.01
YoloV7Tiny	COCO	640x640x3	7.01	6.24	GPL-3.0	mAP	37.289		36.953			37.098			279	256.12
YoloV8L	COCO	640x640x3	85.13	43.69	AGPL-3.0	mAP	52.573		51.681			51.681			82	27.52
YoloV8M	COCO	640x640x3	41.13	25.91	AGPL-3.0	mAP	50.111		48.78			49.225			123	52.1
YoloV8N	COCO	640x640x3	4.89	3.18	AGPL-3.0	mAP	37.316		36.381			36.628			294	330.24
YoloV8S	COCO	640x640x3	15.24	11.18	AGPL-3.0	mAP	44.803		43.845			44.119			239	137.34
YoloV8X	COCO	640x640x3	132.08	68.23	AGPL-3.0	mAP	53.635		52.716			-	-	-	47	16.46
YoloV9C	COCO	640x640x3	53.92	25.31	GPL-3.0	mAP	52.160		49.748			45.475			82	40.84
YoloV9S	COCO	640x640x3	14.50	7.13	GPL-3.0	mAP	45.990		44.255			44.895			258	138.4
YoloV9T	COCO	640x640x3	4.56	2.03	GPL-3.0	mAP	37.705		34.807			36.745			312	317
YoloXLLeaky	COCO	640x640x3	78.01	54.17	Apache-2.0	mAP	48.623		48.311			-	-	-	109	972.18
YoloXS	COCO	640x640x3	14.41	8.96	Apache-2.0	mAP	40.290		40.057			40.107			315	153.33
YoloXSLeaky	COCO	640x640x3	13.49	8.96	Apache-2.0	mAP	38.298		37.814			38.028			314	153.73
YoloXSWideLeaky	COCO	640x640x3	29.89	20.12	Apache-2.0	mAP	42.636		42.366			42.471			207	70.24
YoloXTiny	COCO	416x416x3	3.55	5.05	Apache-2.0	mAP	32.605		32.349			32.465			820	548.39

Face Detection

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
SCRFD10G	WiderFace	640x640x3	13.41	4.23	Apache-2.0	AP(Easy) AP(Med) AP(Hard)	95.469 94.021 82.674		95.44 94.036 82.547			95.448			266	141.07
SCRFD2_5G	WiderFace	640x640x3	3.46	817.96	Apache-2.0	AP(Easy) AP(Med) AP(Hard)	93.888 92.042 77.000		93.747 92.05 76.81			-	-	-	315	367.38
SCRFD500M	WiderFace	640x640x3	764.58	626.34	Apache-2.0	AP(Easy) AP(Med) AP(Hard)	91.080 88.467 69.375		91.047 88.423 69.139			90.941			536	932.04
YOLOv5m_Face	WiderFace	640x640x3	25.84	21.04	MIT	AP(Easy) AP(Med) AP(Hard)	95.507 94.027 85.649		95.53 94.078 86.338			95.675			204	80.18
YOLOv5s_Face	WiderFace	640x640x3	8.53	7.06	MIT	AP(Easy) AP(Med) AP(Hard)	94.570 92.940 83.698		94.738 93.124 84.367			94.736			358	243.85
YOLOv7_Face	WiderFace	640x640x3	54.63	36.56	MIT	AP(Easy) AP(Med) AP(Hard)	96.925 95.689 88.337		96.954 95.67 88.267			97.008			120	42.34
YOLOv7_W6_Face	WiderFace	960x960x3	100.22	69.90	MIT	AP(Easy) AP(Med) AP(Hard)	96.410 95.091 88.610		96.439 95.176 88.77			96.473			68	22.83
YOLOv7_W6_TTA_Face	WiderFace	1280x1280x3	178.16	69.90	MIT	AP(Easy) AP(Med) AP(Hard)	95.890 94.929 89.952		95.993 95.027 90.201			-	-	-	37	12.67
YOLOv7s_Face	WiderFace	640x640x3	9.35	4.27	MIT	AP(Easy) AP(Med) AP(Hard)	94.860 93.300 85.304		94.828 93.223 85.147			94.985			275	203.04

Image De-noising

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
DnCNN_15	BSD68	512x512x1	145.79	555.14	MIT	PSNR SSIM	31.709 0.8905		31.5018 0.8881			-	-	-	35	14.15
DnCNN_25	BSD68	512x512x1	145.79	555.14	MIT	PSNR SSIM	29.1919 0.8276		28.7069 0.8157			-	-	-	35	14.16
DnCNN_50	BSD68	512x512x1	145.79	555.14	MIT	PSNR SSIM	26.1882 0.7184		24.8498 0.6711			-	-	-	35	13.76

Depth Estimation

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
FastDepth	NYU	224x224x3	547.19	1.38	MIT	RMSE	0.604		0.647			-	-	-	512	1536.02

Pose Estimation

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
YOLOV8M_Pose	COCOPose	640x640x3	42.18	26.49	AGPL-3.0	mAP	63.196		61.628			62.088			120	51.11
YOLOV8S_Pose	COCOPose	640x640x3	16.05	11.66	AGPL-3.0	mAP	58.342		57.509			57.923			186	137.35

Semantic Segmentation

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
BiSeNetV1	Cityscapes	1024x2048x3	118.98	13.27	MIT	mIoU	75.367		75.104			-	-	-	19	15.05
BiSeNetV2	Cityscapes	1024x2048x3	99.14	3.35	MIT	mIoU	74.951		74.805			74.771			28	19.28
DeepLabV3PlusMobilenet	VOCSegmentation	512x512x3	26.62	5.80	MIT	mIoU	70.806		67.986			68.879			243	81.53

Instance Segmentation

Class Name	Dataset	Input Resolution	Operations (MFLOPs)	Parameters (M)	License	Metric	Original (FP32)		Quantized (INT8)
							Original (FP32)		Q-Lite			Q-Pro			Performance
							Accuracy	ONNX	Accuracy	DXNN	JSON	Accuracy	DXNN	JSON	FPS	FPS/Watt
YoloV5L_Seg	COCO	640x640x3	76.77	47.89	AGPL-3.0	mAP	39.293		39.245			39.301			102	30.02
YoloV5M_Seg	COCO	640x640x3	37.29	21.97	AGPL-3.0	mAP	36.061		36.018			-	-	-	132	57.13
YoloV5N_Seg	COCO	640x640x3	4.11	1.99	AGPL-3.0	mAP	22.866		22.4			22.691			179	342.84
YoloV5S_Seg	COCO	640x640x3	14.23	7.61	AGPL-3.0	mAP	31.079		30.729			30.967			161	141.65
YoloV8N_Seg	COCO	640x640x3	6.95	3.40	AGPL-3.0	mAP	29.775		29.588			29.81			172	226.86
YoloV8S_Seg	COCO	640x640x3	22.43	11.84	AGPL-3.0	mAP	36.044		35.892			35.996			142	85.5
YoloV8M_Seg	COCO	640x640x3	57.03	27.29	AGPL-3.0	mAP	39.683		39.296			-	-	-	95	37.11

Test Environment

Hardware Info: Intel(R) Core(TM) i5-14600K, 32G RAM
SDK Version: dx-com v2.2.0, dx-rt v3.2.0
Benchmark Cmd: run_model -m <MODEL_FILE> --use-ort -t 5 -b

* Performance results may vary depending on the specific hardware configuration.