BIP3D: Bridging 2D Images and 3D Perception for Embodied Intelligence

Xuewu Lin, Tianwei Lin, Lichao Huang, Hongyu Xie, Zhizhong Su

The Architecture Diagram of BIP3D, where the red stars indicate the parts that have been modified or added compared to the base model, GroundingDINO, and dashed lines indicate optional elements.

Results on EmbodiedScan Benchmark

We made several improvements based on the original paper, achieving better 3D perception results. The main improvements include the following two points:

New Fusion Operation: We enhanced the decoder by replacing the deformable aggregation (DAG) with a 3D deformable attention mechanism (DAT). Specifically, we improved the feature sampling process by transitioning from bilinear interpolation to trilinear interpolation, which leverages depth distribution for more accurate feature extraction.
Mixed Data Training: To optimize the grounding model's performance, we adopted a mixed-data training strategy by integrating detection data with grounding data during the grounding finetuning process.

1. Results on Multi-view 3D Detection Validation Dataset

Model	Inputs	Op	Overall	Head	Common	Tail	Small	Medium	Large	ScanNet	3RScan	MP3D	ckpt	log
BIP3D	RGB	DAG	16.57	23.29	13.84	12.29	2.67	17.85	12.89	19.71	26.76	8.50	-	-
BIP3D	RGB	DAT	16.67	22.41	14.19	13.18	3.32	17.25	14.89	20.80	24.18	9.91	-	-
BIP3D	RGB-D	DAG	22.53	28.89	20.51	17.83	6.95	24.21	15.46	24.77	35.29	10.34	-	-
BIP3D	RGB-D	DAT	23.24	31.51	20.20	17.62	7.31	24.09	15.82	26.35	36.29	11.44	-	-

2. Results on Multi-view 3D Grounding Mini Dataset

Model	Inputs	Op	Overall	Easy	Hard	View-dep	View-indep	ScanNet	3RScan	MP3D	ckpt	log
BIP3D	RGB	DAG	44.00	44.39	39.56	46.05	42.92	48.62	42.47	36.40	-	-
BIP3D	RGB	DAT	44.43	44.74	41.02	45.17	44.04	49.70	41.81	37.28	-	-
BIP3D	RGB-D	DAG	45.79	46.22	40.91	45.93	45.71	48.94	46.61	37.36	-	-
BIP3D	RGB-D	DAT	58.47	59.02	52.23	60.20	57.56	66.63	54.79	46.72	-	-

3. Results on Multi-view 3D Grounding Validation Dataset

Model	Inputs	Op	Mixed Data	Overall	Easy	Hard	View-dep	View-indep	ScanNet	3RScan	MP3D	ckpt	log
BIP3D	RGB	DAG	No	45.81	46.21	41.34	47.07	45.09	50.40	47.53	32.97	-	-
BIP3D	RGB	DAT	No	47.29	47.82	41.42	48.58	46.56	52.74	47.85	34.60	-	-
BIP3D	RGB-D	DAG	No	53.75	53.87	52.43	55.21	52.93	60.05	54.92	38.20	-	-
BIP3D	RGB-D	DAT	No	61.36	61.88	55.58	62.43	60.76	66.96	62.75	46.92	-	-
BIP3D	RGB-D	DAT	Yes	66.58	66.99	62.07	67.95	65.81	72.43	68.26	51.14	-	-

4. Results on Multi-view 3D Grounding Test Dataset

Model	Overall	Easy	Hard	View-dep	View-indep	ckpt	log
EmbodiedScan	39.67	40.52	30.24	39.05	39.94	-	-
SAG3D*	46.92	47.72	38.03	46.31	47.18	-	-
DenseG*	59.59	60.39	50.81	60.50	59.20	-	-
BIP3D	67.38	68.12	59.08	67.88	67.16	-	-
BIP3D-Base	70.53	71.22	62.91	70.69	70.47	-	-

Citation

@article{lin2024bip3d,
  title={BIP3D: Bridging 2D Images and 3D Perception for Embodied Intelligence},
  author={Lin, Xuewu and Lin, Tianwei and Huang, Lichao and Xie, Hongyu and Su, Zhizhong},
  journal={arXiv preprint arXiv:2411.14869},
  year={2024}
}