VAEforPointCloud

VAEforPointCloud/
├── lib/
│   ├── __init__.py            # An empty file required to recognize this directory as a module
│   ├── ChamferDis.py          # Chamfer Distance calculation
│   ├── Dataloader.py          # DataLoader for preparing training data
│   ├── EarthMoversDis.py      # Earth Mover's Distance calculation
│   ├── file_io.py             # File input/output functions for reading OFF files
│   ├── ☆mogvae_v2_models.py  # Mixture of Gaussians VAE model (MoGVAE) for non-flattened input
│   ├── point_cloud.py         # Functions for point cloud visualization and rotation
│   ├── sampling.py            # PointSampler class for point cloud data
│   ├── train.py               # Training function for VAE models
│   ├── utils.py               # Utility functions (memory check, data handling)
│   ├── ☆vae_v2_models.py     # Standard VAE model (standVAE) for non-flattened input
│   ├── ☆vqvae_v2_models.py   # VQ-VAE model for non-flattened input
│   ├── ☆SetVAE.py   # SetVAE model for non-flattened input
│   ├── ☆ISAB.py              # ISAB(Induced set attention block) model for non-flattened input
│   └── visualize_loss.py      # Functions for visualizing training loss and results
|
├── main_for_standardVAE_v2_(MSE).ipynb # Main VAE script for training and evaluation ( Loss function MSE + KL_D ) for non-flattened input
├── main_for_MoGVAE_v2_(MSE).ipynb      # Main MoG-VAE script for training and evaluation ( Loss function MSE + KL_D ) for non-flattened input
├── main_for_MoGVAE_v2_(MSE+CD).ipynb   # Main MoG-VAE script for training and evaluation ( Loss function MSE + CD+ KL_D ) for non-flattened input
├── main_for_SetVAE_v2_(MSE).ipynb      # Main SetVAE w/ ISAB( Induced set attention block) script for training and evaluation ( Loss function MSE + KL_D ) for non-flattened input
├── main_for_VQVAE_v2_(MSE).ipynb       # Main VQ-VAE script for training and evaluation ( Loss function MSE ) for non-flattened input
├── requirements.txt                    # List of required Python packages
└── README.md                           # Project description

Architecture	VAE	MoG-VAE	MoG-VAE	SetVAE	SetVAE	VQ-VAE	ISAB+VQ-VAE	MAB+VQ-VAE	ISAB+SoftVQ-VAE
Encoder	Pointwise conv + Max Pooling	Same as Left	Same as Left	Induced Set Attention Block	Same as Left	Pointwise conv + Max Pooling	Induced Set Attention Block	Multihead Attention Block	Induced Set Attention Block
Decoder	Deconvolution (Transpose Conv)	Same as Left	Same as Left	Same as Left	Same as Left	Same as Left	Same as Left	Same as Left	Same as Left
Loss	MSE + KLD	MSE + KLD	MSE + 2×CD + KLD	MSE + 0.8 × KLD	0.5×MSE + 10×CD + 0.4×KLD	MSE + Codebook + Commitment	Same as Left	Same as Left	MSE + Codebook (SoftVQ) + Commitment
Learning Rate	1.0E-05	1.0E-05	1.0E-04	5.0E-05	1.0E-04	1.0E-03	1.0E-04	1.0E-04	1.0E-04
CD
Design 1	0.0245	0.0247	0.0145	0.0239	0.0166	0.0188	0.0116	0.0133	0.0014
Design 2	0.0247	0.0226	0.0154	0.0187	0.0172	0.0208	0.0112	0.0046	0.0005
Design 3	0.0390	0.0161	0.0231	0.0188	0.0226	0.0105	0.0051	0.0092	0.0006
Design 4	0.0303	0.0227	0.0287	0.0210	0.0156	0.0130	0.0035	0.0033	0.0005
Design 5	0.0333	0.0287	0.0286	0.0343	0.0223	0.0188	0.0094	0.0070	0.0009
Design 6	0.0292	0.0174	0.0218	0.0185	0.0131	0.0121	0.0087	0.0052	0.0004
Design 7	0.0463	0.0277	0.0198	0.0186	0.0205	0.0106	0.0119	0.0208	0.0004
Design 8	0.0286	0.0281	0.0236	0.0151	0.0172	0.0135	0.0125	0.0052	0.0006
Design 9	0.0315	0.0316	0.0221	0.0180	0.0349	0.0258	0.0166	0.0068	0.0007
Average	0.0319	0.0244	0.0220	0.0208	0.0200	0.0160	0.0101	0.0084	0.0007
EMD
Design 1	0.0167	0.0181	0.0088	0.0210	0.0112	0.0135	0.0068	0.0084	0.0007
Design 2	0.0145	0.0131	0.0082	0.0103	0.0093	0.0116	0.0058	0.0023	0.0002
Design 3	0.0298	0.0088	0.0142	0.0105	0.0140	0.0055	0.0026	0.0047	0.0003
Design 4	0.0195	0.0136	0.0189	0.0119	0.0086	0.0069	0.0018	0.0017	0.0002
Design 5	0.0217	0.0176	0.0177	0.0239	0.0132	0.0104	0.0048	0.0035	0.0004
Design 6	0.0267	0.0111	0.0140	0.0131	0.0079	0.0070	0.0047	0.0027	0.0002
Design 7	0.0400	0.0176	0.0112	0.0103	0.0118	0.0054	0.0062	0.0117	0.0002
Design 8	0.0213	0.0200	0.0161	0.0086	0.0103	0.0076	0.0069	0.0027	0.0003
Design 9	0.0197	0.0200	0.0130	0.0100	0.0238	0.0157	0.0091	0.0035	0.0003
Average	0.0233	0.0155	0.0136	0.0133	0.0122	0.0093	0.0054	0.0046	0.0003