Efficient Models

Model	Parameters	GMacs	GFLOPS	F-Pass Memory	Loss	IoU	Epoch	Optimizer	Criterion	LR	Dataset	Paper	Code	Type
HarDNet-DWS-39	2.2M	0.44	0.88	29.77MB	0	0	0	N	N	N	38-Cloud	Link	Code	Classification
FCHarDNet	453K	0.85	1.7	22.04MB	0.106	0	289	Adam	DiceLoss	0.1e-4	38-Cloud	Link	Code	Segmentation
FCHarDNet	453K	0.85	1.7	22.04MB	0.049	0	943	Adam	DiceLoss	0.1e-4	38-Cloud	Link	Code	Segmentation
ENet	349K	0.4	0.8	145MB	0.038	0	982	Adam	BCELoss	.1e-4	38-Cloud	Link	Code	Segmentation
SegNet	24.4M	30.73	64.4	343.77MB	-0.250	0	281	Adam	BCELoss	.1e-4	38-Cloud	Link	Code	Segmentation
SegNet	24.4M	30.73	64.4	343.77MB	0.532	0	296	SGD	BCELoss	.1e-4	38-Cloud	Link	Code	Segmentation
UNet	31M	41.9	83.8	419.95MB	0.054	0	293	Adam	DiceLoss	.1e-4	38-Cloud	Link	Code	Segmentation
FineTuned-FCHarDNet-Cloud38-RSI	453K	0.85	1.7	22.04MB	0.142	0	290	Adam	DiceLoss	0.1e-4	38-Cloud - RSI	Link	Code	Segmentation
ENet	349K	0.4	0.8	145MB	0.22	0	286	Adam	BCELoss	.1e-4	38-Cloud	Link	Code	Segmentation

Top Models Trained in CamVid Dataset

Model	mIoU	Sky	Road	Building	Sidewalk	Car	Tree	Fence	Cyclist	Pedestrian	Pole	Sign
FCHarDNet	0.506	51.95	50.99	40.20	39.32	37.69	35.33	19.1	16.18	13.35	10.02	8.45
ENet	0.566	52.91	52.59	45.32	45.02	40.52	42.20	23.94	20.34	19.70	18.58	0