6.1.4. ModelZoo

6.1.4.1. Classification

network	float	qat	quantization	dataset
VargNet	73.94	73.64	73.58	ImageNet
VargConvNet	78.98	78.90	78.89	ImageNet
efficientnasnetm	80.25	79.97	79.96	ImageNet
efficientnasnets	76.64	75.96	75.94	ImageNet
EfficientNet	74.31	74.23	74.07	ImageNet
MobileNetV1	74.10	73.59	73.62	ImageNet
ResNet18	72.04	72.01	72.02	ImageNet

6.1.4.2. Detection

FCOS

network	backbone	float	qat	quantization	dataset
FCOS-EfficientNetB0	EfficientNet-B0	36.27	35.62	35.55	MS COCO
FCOS-EfficientNetB1	EfficientNet-B1	41.36	40.84	40.75	MS COCO
FCOS-EfficientNetB2	EfficientNet-B2	45.35	45.11	45.08	MS COCO
FCOS-EfficientNetB3	EfficientNet-B3	48.02	47.58	47.56	MS COCO

6.1.4.3. Segmentation

| network | backbone | float | qat | quantization | dataset | | UNet | MobileNetV1 | 68.02 | 67.56 | 67.53 | Cityscapes | | Deeplab | EfficientNet-M0 | 76.30 | 76.14 | 76.13 | Cityscapes | | Deeplab | EfficientNet-M1 | 77.94 | 77.66 | 77.65 | Cityscapes | | Deeplab | EfficientNet-M2 | 78.82 | 78.62 | 78.63 | Cityscapes | | FastScnn | EfficientNet-B0lite | 69.97 | 69.62 | 69.72 | Cityscapes |

6.1.4.4. Lidar

PointPillars

network	backbone	float	qat	quantization	dataset
PointPillars	SequentialBottleNeck	74.04	68.58	68.25	KITTI3D