Building a Real-Time 2D Lidar Using Deep Learning

Table 1

Results on KITTI test dataset.

Small is better						Large is better
Network	Params	Seq Rel	Abs Rel	Scale invar.	RMSL	RMS	EVAR				FPS
	—	0.076	0.465	0.105	0.468	0.141	0.000	0.872	0.692	0.290	—
VGG19	58.1 M	0.016	0.152	0.029	0.240	0.079	0.693	0.974	0.926	0.777	12
Proposed	3.2 M	0.018	0.176	0.031	0.251	0.082	0.672	0.976	0.916	0.724	40
DORN	96.8 M	0.020	0.172	0.033	0.256	0.085	0.617	0.969	0.920	0.750	6
ResNet50	23.7 M	0.021	0.180	0.035	0.266	0.085	0.636	0.966	0.907	0.733	21
FCRN	33.7 M	0.025	0.201	0.053	0.324	0.092	0.577	0.957	0.888	0.702	10