Disentangled Representation Learning on Cars3D
91FactorVAESlowVAE
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SlowVAEData=LAP, Aggregation Statistic=Median2020.07 | 91 | 100 | 9.7 | 51 | 94.4 | 1.7 | — | — | — | — | — | |
| SlowVAEData=LAP-NC, Aggregation Statistic=Mean2020.07 | 90.9 | 100 | 9.5 | 50.2 | 95 | 1.7 | — | — | — | — | — | |
| SlowVAEData=LAP-NC, Aggregation Statistic=Median2020.07 | 90.8 | 100 | 9.3 | 50 | 94.6 | 0.9 | — | — | — | — | — | |
| SlowVAEData=UNI, Aggregation Statistic=Median2020.07 | 90.4 | 100 | 15.7 | 48.9 | 95.7 | 1.6 | — | — | — | — | — | |
| SlowVAEData=UNI, Aggregation Statistic=Mean2020.07 | 90.4 | 100 | 15.4 | 48 | 95.4 | 1.6 | — | — | — | — | — | |
| Ada-GVAEData=LOC, Aggregation Statistic=Median2020.07 | 90.2 | 100 | 15 | 54 | 93.9 | 9.4 | — | — | — | — | — | |
| SlowVAEData=LAP, Aggregation Statistic=Mean2020.07 | 90.2 | 100 | 10.4 | 50.9 | 94.1 | 2 | — | — | — | — | — | |
| β-VAEData=i.i.d., Aggregation Statistic=Median2020.07 | 87.9 | 100 | 8.8 | 22.5 | 90.2 | 1 | — | — | — | — | — | |
| Ada-ML-VAEData=LOC, Aggregation Statistic=Median2020.07 | 87.4 | 100 | 14.7 | 45.6 | 94.6 | 2.8 | — | — | — | — | — | |
| STEERINGDRLN=202026.07 | 1 | — | — | 0.624 | — | — | — | — | — | — | — | |
| Soft TPRRepresentation Type=Fully continuous compositional representations2024.12 | 0.999 | — | — | 0.863 | — | — | — | — | — | — | — | |
| STEERINGDRLN=102026.07 | 0.987 | — | — | 0.527 | — | — | — | — | — | — | — | |
| SCFM (β-TCVAE)Category=Flow, Endpoint Regularizer=β-TCVAE2026.05 | 0.977 | — | — | 0.357 | — | — | — | — | — | — | — | |
| DisDiffMethod Category=Diffusion-based, Representation Type=vector-valued, Resolution=64x642024.02 | 0.976 | — | — | 0.232 | — | — | — | — | — | — | — | |
| DisDiffCategory=Diffusion2026.05 | 0.976 | — | — | 0.232 | — | — | — | — | — | — | — | |
| DisDiffN=202026.07 | 0.976 | — | — | 0.232 | — | — | — | — | — | — | — | |
| VCTType=Concept-based2022.05 | 0.966 | — | — | 0.382 | — | — | — | — | — | — | — | |
| VCTRepresentation Type=Symbolic vector-tokened compositional representations2024.12 | 0.966 | — | — | 0.382 | — | — | — | — | — | — | — | |
| EncDiffCategory=Diffusion2026.05 | 0.948 | — | — | 0.357 | — | — | — | — | — | — | — | |
| Ada-GVAE-kRepresentation Type=Symbolic scalar-tokened compositional representations2024.12 | 0.947 | — | — | 0.664 | — | — | — | — | — | — | — | |
| Generalized GBN2024.08 | 0.946 | 1 | — | — | 0.897 | 0.023 | — | — | 0.323 | 0.813 | 0.218 | |
| DyGAN=202026.07 | 0.941 | — | — | 0.414 | — | — | — | — | — | — | — | |
| SCFM (β-VAE)Category=Flow, Endpoint Regularizer=β-VAE2026.05 | 0.94 | — | — | 0.337 | — | — | — | — | — | — | — | |
| GANspaceType=Pre-trained GAN-based2022.05 | 0.932 | — | — | 0.209 | — | — | — | — | — | — | — | |
| GSMethod Category=Pre-trained GAN-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.932 | — | — | 0.209 | — | — | — | — | — | — | — | |
| BetaTCVAE2024.08 | 0.929 | 1 | — | — | 0.931 | 0.014 | — | — | 0.318 | 0.817 | 0.246 | |
| β-VAE2024.08 | 0.923 | 1 | — | — | 0.93 | 0.005 | — | — | 0.316 | 0.559 | 0.224 | |
| FDAEN=202026.07 | 0.918 | — | — | 0.232 | — | — | — | — | — | — | — | |
| FDAECategory=Diffusion2026.05 | 0.912 | — | — | 0.329 | — | — | — | — | — | — | — | |
| Factor VAE2024.08 | 0.907 | 1 | — | — | 0.889 | 0.009 | — | — | 0.144 | 0.682 | 0.142 | |
| FactorVAEType=VAE-based2022.05 | 0.906 | — | — | 0.161 | — | — | — | — | — | — | — | |
| FactorVAEMethod Category=VAE-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.906 | — | — | 0.161 | — | — | — | — | — | — | — | |
| Slow-VAERepresentation Type=Symbolic scalar-tokened compositional representations2024.12 | 0.902 | — | — | 0.509 | — | — | — | — | — | — | — | |
| β-VAECategory=VAE2026.05 | 0.887 | — | — | 0.218 | — | — | — | — | — | — | — | |
| GVAERepresentation Type=Symbolic scalar-tokened compositional representations2024.12 | 0.877 | — | — | 0.262 | — | — | — | — | — | — | — | |
| Modular IBN=202026.07 | 0.877 | — | — | 0.365 | — | — | — | — | — | — | — | |
| ClosedFormType=Pre-trained GAN-based2022.05 | 0.873 | — | — | 0.243 | — | — | — | — | — | — | — | |
| DIP-VAE2024.08 | 0.873 | 1 | — | — | 0.837 | 0.014 | — | — | 0.261 | 0.704 | 0.142 | |
| DeepSpectralType=Pre-trained GAN-based2022.05 | 0.871 | — | — | 0.222 | — | — | — | — | — | — | — | |
| ML-VAERepresentation Type=Symbolic scalar-tokened compositional representations2024.12 | 0.87 | — | — | 0.216 | — | — | — | — | — | — | — | |
| Beta-TCVAEType=VAE-based2022.05 | 0.855 | — | — | 0.14 | — | — | — | — | — | — | — | |
| DisCoType=Pre-trained GAN-based2022.05 | 0.855 | — | — | 0.271 | — | — | — | — | — | — | — | |
| beta-TCVAEMethod Category=VAE-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.855 | — | — | 0.14 | — | — | — | — | — | — | — | |
| DisCoMethod Category=Pre-trained GAN-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.855 | — | — | 0.271 | — | — | — | — | — | — | — | |
| β-TCVAECategory=VAE2026.05 | 0.855 | — | — | 0.14 | — | — | — | — | — | — | — | |
| LatentDiscoveryType=Pre-trained GAN-based2022.05 | 0.852 | — | — | 0.216 | — | — | — | — | — | — | — | |
| LDMethod Category=Pre-trained GAN-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.852 | — | — | 0.216 | — | — | — | — | — | — | — | |
| DyGACategory=Diffusion2026.05 | 0.846 | — | — | 0.307 | — | — | — | — | — | — | — | |
| EncDiffMethod Category=Diffusion-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.773 | — | — | 0.279 | — | — | — | — | — | — | — | |
| EncDiffN=202026.07 | 0.765 | — | — | 0.258 | — | — | — | — | — | — | — | |
| VAE2024.08 | 0.753 | 0.999 | — | — | 0.836 | 0.027 | — | — | 0.11 | 0.702 | 0.07 | |
| EncDiffN=102026.07 | 0.714 | — | — | 0.208 | — | — | — | — | — | — | — | |
| ShuRepresentation Type=Symbolic scalar-tokened compositional representations2024.12 | 0.573 | — | — | 0.032 | — | — | — | — | — | — | — | |
| InfoGAN-CRType=GAN-based2022.05 | 0.411 | — | — | 0.02 | — | — | — | — | — | — | — | |
| InfoGAN-CRMethod Category=GAN-based, Representation Type=scalar-valued, Resolution=64x642024.02 | 0.411 | — | — | 0.02 | — | — | — | — | — | — | — | |
| COMETType=Concept-based2022.05 | 0.339 | — | — | 0.024 | — | — | — | — | — | — | — | |
| COMETRepresentation Type=Symbolic vector-tokened compositional representations2024.12 | 0.339 | — | — | 0.024 | — | — | — | — | — | — | — | |
| Ada-GVAE-kRepresentation type=Symbolic scalar-tokened, Learnable parameter control=false2024.12 | — | 1 | 0.395 | — | — | — | — | — | — | — | — | |
| Ada-GVAE-k*Representation type=Symbolic scalar-tokened, Learnable parameter control=true2024.12 | — | 1 | 0.321 | — | — | — | — | — | — | — | — | |
| CFASL2026.05 | — | — | — | — | — | — | 66.88 | 88 | — | — | — | |
| CLG-VAE2026.05 | — | — | — | — | — | — | 67.58 | 91 | — | — | — | |
| COMETRepresentation type=Symbolic vector-tokened2024.12 | — | 0.343 | 0 | — | — | — | — | — | — | — | — | |
| GVAERepresentation type=Symbolic scalar-tokened, Learnable parameter control=false2024.12 | — | 1 | 0.096 | — | — | — | — | — | — | — | — | |
| GVAE*Representation type=Symbolic scalar-tokened, Learnable parameter control=true2024.12 | — | 1 | 0.1 | — | — | — | — | — | — | — | — | |
| LG-VAE2026.05 | — | — | — | — | — | — | 87.1 | 88 | — | — | — | |
| ML-VAERepresentation type=Symbolic scalar-tokened, Learnable parameter control=false2024.12 | — | 1 | 0.088 | — | — | — | — | — | — | — | — | |
| ML-VAE*Representation type=Symbolic scalar-tokened, Learnable parameter control=true2024.12 | — | 1 | 0.05 | — | — | — | — | — | — | — | — | |
| OursTraining Phase=Phase 1 only, Training Duration (Epochs)=P1 epochs2026.05 | — | — | — | — | — | — | 67.96 | 92 | — | — | — | |
| OursTraining Phase=Phase 1 only, Training Duration (Epochs)=P1+P2 epochs2026.05 | — | — | — | — | — | — | 58.17 | 87 | — | — | — | |
| OursTraining Phase=Phase 1 + Phase 2, Training Duration (Epochs)=P1+P2 epochs2026.05 | — | — | — | — | — | — | 58.44 | 95 | — | — | — | |
| ShuRepresentation type=Symbolic scalar-tokened, Learnable parameter control=false2024.12 | — | 0.912 | 0.025 | — | — | — | — | — | — | — | — | |
| Shu*Representation type=Symbolic scalar-tokened, Learnable parameter control=true2024.12 | — | 0.923 | 0.015 | — | — | — | — | — | — | — | — | |
| Slow-VAERepresentation type=Symbolic scalar-tokened, Learnable parameter control=false2024.12 | — | 1 | 0.104 | — | — | — | — | — | — | — | — | |
| Slow-VAE*Representation type=Symbolic scalar-tokened, Learnable parameter control=true2024.12 | — | 1 | 0.071 | — | — | — | — | — | — | — | — | |
| Soft TPRRepresentation type=Fully continuous2024.12 | — | 1 | 0.348 | — | — | — | — | — | — | — | — | |
| VCTRepresentation type=Symbolic vector-tokened2024.12 | — | 1 | 0.117 | — | — | — | — | — | — | — | — | |
| β-VAE2026.05 | — | — | — | — | — | — | 59.09 | 85 | — | — | — |