List your company
ResearchFormerly WizWand

3D Object Recognition on ShapeNet

99.59AccuracyOurs w/ PointBind
12.084434.802257.5280.2378Apr 28, 2023Jul 6, 2023Sep 13, 2023Nov 22, 2023Jan 30, 2024Apr 8, 2024Jun 17, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Ours w/ PointBindSetting=Fine-tuning2024.0399.59
Meta-TransformerSetting=Fine-tuning2024.0399.3
PointBindSetting=Fine-tuning, Protocol=linear2024.0399.09
PointBind (+Event)Setting=Fine-tuning2024.0399.09
Ours w/ PointBindSetting=Zero-shot2024.0398.96
PointBindSetting=Zero-shot2024.0398.85
PointBind (+Event)Setting=Zero-shot2024.0398.85
LLaVA-vicuna-13bModality=Image (Multi-view)2024.0673.45
LLaNA-7bModality=NeRF2024.0667.14
LLaVA-vicuna-13bModality=Image (Front-view)2024.0666.13
LLaVA-vicuna-13bModality=Image (Back-view)2024.0663.9
BLIP-2 FlanT5-xxlModality=Image (Front-view)2024.0663.67
BLIP-2 FlanT5-xxlModality=Image (Back-view)2024.0661.47
MetaGPBackbone=DGCNN, # Classes=16, Reusing Method=MetaGP2023.0460.9
3D-LLMModality=Mesh + Multi-view2024.0660.55
LLaVA-vicuna-7bModality=Image (Front-view)2024.0660.25
LLaVA-vicuna-7bModality=Image (Back-view)2024.0657
PointLLM-7bModality=Point cloud2024.0650.14
GPT4Point-Opt-2.7bModality=Point cloud2024.0641.93
Vanilla ReusingBackbone=DGCNN, # Classes=16, Reusing Method=Vanilla2023.0415.45