Estimation method of category-level multi-object rigid body 6D pose

Methods	Data type	Ape	Can	Cat	Drill	Duck	Eggb.	Glue	Hol.	Avg.
DPOD^［12］	syn.	2.3	4.0	1.2	10.5	7.2	4.4	12.9	7.5	6.3
CDPN^［8］	syn.	20.0	15.1	16.4	5.0	22.2	36.1	27.9	24.0	20.8
DSC-PoseNet^［20］	pbr+RGB	9.1	21.1	26.0	33.5	12.2	39.4	37.0	20.4	24.8
PyraPose^［16］	pbr	18.5	46.4	11.7	48.2	19.4	16.7	30.7	33.0	28.1
OSOP^［13］	-	-	-	-	-	-	-	-	-	31.2
Self6D^［21］	pbr+RGBD	13.7	43.2	18.7	32.5	14.4	57.8	52.3	22.0	32.1
DAKDN^［22］	pbr+RGB	-	-	-	-	-	-	-	-	33.7
SD-Pose^［23］	pbr	21.5	56.7	17.0	44.4	27.6	42.8	45.2	21.6	34.6
CASAPose^［24］	pbr	24.3	59.5	15.2	57.5	26.0	14.7	55.4	34.3	35.9
Ours	pbr	29.2	62.4	18.8	56.1	32.6	36.4	59.8	26.2	40.2

Table 2. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM dataset

View table

View in Article

Table 2. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM dataset

Methods	Ape	Bv.	Cam	Can	Cat	Drill	Duck	Eggb	Glue	Hol.	Iron	Lamp.	Ph.	Avg.
Self6D-LB^［21］	37.2	66.9	17.9	50.4	33.7	47.4	18.3	64.8	59.9	5.2	68.0	35.3	36.5	40.1
PyraPose^［16］	19.9	68.6	49.1	71.5	49.1	61.8	36.3	74.1	72.0	38.1	72.1	54.1	51.9	55.3
SD-Pose^［23］	47.1	66.7	43.8	70.7	62.0	56.0	47.1	82.0	80.8	21.0	67.2	72.1	46.9	58.8
CASAPose^［24］	32.1	69.4	57.4	64.8	49.1	70.3	36.3	73.1	64.3	48.2	73.4	71.9	62.5	59.4
Ours	31.9	70.3	57.8	67.0	47.1	71.0	37.2	72.6	67.2	48.8	73.8	76.2	64.6	60.4

Table 3. Comparison of 2DP and ADD/S-Recall indexes of different target number experiments on LM-O and LM datasets
View table
View in Article
Table 3. Comparison of 2DP and ADD/S-Recall indexes of different target number experiments on LM-O and LM datasets
Target quantity LM-O LM
2DP ADD/S 2DP ADD/S
13 Obj. 51.4 37.4 79.4 60.4
8 Obj. 53.6 40.2 80.0 58.9
4 Obj.（×2） 57.6 42.7 81.0 60.8

Table 4. Comparison of ADD/S-Recall indexes of different module experiments on LM-O and LM data sets
View table
View in Article
Table 4. Comparison of ADD/S-Recall indexes of different module experiments on LM-O and LM data sets
CABD FOSM MFED LM-O LM
2DP ADD/S 2DP ADD/S
√ 53.0 36.0 75.5 58.0
√ 54.6 37.5 79.2 59.9
√ 55.6 39.9 78.8 60.2
√ √ √ 55.7 40.2 79.4 60.4

Tools

Get Citation

Copy Citation Text

Shuo CHENG, Di JIA, Liu YANG, Dekun HE. Estimation method of category-level multi-object rigid body 6D pose[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(3): 457

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Jun. 26, 2024

Accepted: --

Published Online: Apr. 27, 2025

The Author Email: Shuo CHENG (lntu_cs@163.com)

DOI:10.37188/CJLCD.2024-0182

Topics

Table 1. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM-O dataset

Table 1. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM-O dataset

Table 2. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM dataset

Table 2. ADD/S-Recall index and performance in this paper compared with the most advanced methods on the LM dataset

Table 3. Comparison of 2DP and ADD/S-Recall indexes of different target number experiments on LM-O and LM datasets

Table 3. Comparison of 2DP and ADD/S-Recall indexes of different target number experiments on LM-O and LM datasets

Table 4. Comparison of ADD/S-Recall indexes of different module experiments on LM-O and LM data sets

Table 4. Comparison of ADD/S-Recall indexes of different module experiments on LM-O and LM data sets