Enhancing Point Features with Spatial Information for Point-Based 3D Object Detection

Table 2

Quantitative comparison with advanced methods on the SUN-RGBD test set.

Method	Modality	Bathtub	Bed	Bookshelf	Chair	Desk	Dresser	Nightstand	Sofa	Table	Toilet	3D mAP
PointFusion [35]	L & I	37.3	68.6	37.7	55.1	17.2	24.0	32.3	53.8	31.0	83.8	44.1
F-PointNet [12]	L & I	43.3	81.1	33.3	64.2	24.7	32.0	58.1	61.1	51.1	90.9	54.0
VoteNet [28]	L	74.4	83.0	28.8	75.3	22.0	29.8	62.2	64.0	47.3	90.1	57.7
EPNet [15]	L & I	75.4	85.2	35.4	75.0	26.1	31.3	62.0	67.2	52.1	88.2	59.8
MBDF-Net [36]	L & I	81.5	84.7	33.0	77.3	31.2	29.0	57.7	65.6	49.9	85.5	59.5
Our	L & I	75.6	85.4	35.5	75.6	26.4	31.6	62.5	67.7	52.8	88.6	60.2

L and I represent the LiDAR point cloud and camera image.