SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2210.12918 · 2022

Unsupervised Object Representation Learning using Translation and Rotation Group Equivariant VAE

arXiv · PDF · Open in the Atlas

Code that ran

10 of 20 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 0 failures of our instrument; 10 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2210.12918.json · badge: /api/ran/2210.12918.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
smlc-nysbc/target-vae canonical (classifier label)
classifier basis not recorded on the link
10 ran · 10 no recorded run
FunctionStatusWhere it lives
cluster_acc Ran
ran on a synthesized input
smlc-nysbc/target-vae/clustering_dsprites.py
code served (permissive licence) · get_code("5260d45bd76e16cc")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy.optimize import linear_sum_assignment

# Create small sample data for clustering
np.random.seed(0)

# y_true: true labels, should be a numpy array with shape (n_samples,)
y_true = np.array([0, 0, 1, 1, 2, 2], dtype=np.int64)

# y_pred: predicted labels, should be a numpy array with shape (n_samples,)
# Using permuted/noisy labels to simulate clustering predictions
y_pred = np.array([1, 1, 2, 2, 0, 0], dtype=np.int64)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
cluster_acc Ran
ran on a synthesized input
smlc-nysbc/target-vae/clustering_mnist.py
code served (permissive licence) · get_code("e9e60431bb497e09")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
from scipy.optimize import linear_sum_assignment

# Set random seed for reproducibility
np.random.seed(0)

# Create small numpy arrays for y_true and y_pred
# y_true: true labels
y_true = np.array([0, 1, 0, 1, 2, 2], dtype=np.int64)

# y_pred: predicted labels (same size, but different values to test clustering)
y_pred = np.array([0, 1, 0, 2, 2, 1], dtype=np.int64)

ARGS = {
    "y_true": y_true,
    "y_pred": y_pred,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
compute_2d_ctf Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/ctf.py
code served (permissive licence) · get_code("20061b753ab107e5")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Construct arguments for compute_2d_ctf
# freqs: array of shape (N, 2) containing frequency coordinates (x, y)
freqs = np.array([
    [0.1, 0.2],
    [0.3, 0.4],
], dtype=np.float32)

# dfu: defocus in u direction (in micrometers)
dfu = 1.5

# dfv: defocus in v direction (in micrometers)
dfv = 1.8

# dfang: defocus angle (in radians)
dfang = 0.5

# volt: accelerating voltage (in kV)
volt = 300.0

# cs: spherical aberration coefficient (in mm)
cs = 2.7

# w: amplitude contrast ratio (0 to 1)
w = 0.1

# bfactor: B-factor for envelope (optional, in Angstrom^2)
bfactor = 100.0

ARGS = {
    "freqs": freqs,
    "dfu": dfu,
    "dfv": dfv,
    "dfang": dfang,
    "volt": volt,
    "cs": cs,
    "w": w,
    "bfactor": bfactor,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
crop Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/image.py
code served (permissive licence) · get_code("69f6f08ec67a0821")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 4, dtype family float, shape (2, 4, 4, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function crops the last two dimensions of a stack tensor
# It takes a stack (tensor with shape where last 2 dims are spatial)
# and a size parameter (int) for the crop size

# Create a test stack tensor with shape (2, 4, 6, 8)
# - batch dimension: 2
# - channel dimension: 4
# - spatial dimensions (last 2): 6x8
stack = torch.randn(2, 4, 6, 8)

# size must be smaller than the spatial dimensions (6 and 8)
# The function will crop symmetrically from center
size = 4

ARGS = {
    "stack": stack,
    "size": size,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 4, 4, 4) — shape (2, 4, 4, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
downsample Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/image.py
code served (permissive licence) · get_code("65871fecaa6f5b52")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small 2D array for testing
# Using distinct dimensions: 8x6 for clear probing
x = np.random.randn(8, 6).astype(np.float32)

# Define ARGS dict with all parameters
ARGS = {
    "x": x,
    "factor": 2,
    "shape": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (4, 3) — shape (4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_latent Ran
ran on a synthesized input
smlc-nysbc/target-vae/clustering_dsprites.py
code served (permissive licence) · get_code("1d38fa41d0b89adb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn
import numpy as np

torch.manual_seed(0)

# Create a minimal encoder model for 'unimodal' case
class UnimodalEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(12, 16)  # input flattened y (2 batch * 6 pixels), output mu and logstd
    
    def forward(self, y):
        # y is (batch, flattened_pixels)
        # return z_mu and z_logstd
        # z_mu: (batch, 8) - [theta, dx1, dx2, content1, content2, content3, content4, content5]
        # z_logstd: (batch, 8)
        out = self.fc(y)
        z_mu = out[:, :8]
        z_logstd = out[:, 8:]
        return z_mu, z_logstd

# Create encoder for 'attention' + 'unimodal' case
class AttentionUnimodalEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(12, 64)
    
    def forward(self, y, device):
        batch_size = y.shape[0]
        # Return: attn, sampled_attn, theta_vals, z_vals
        # attn: (batch, 1, attn_dim, attn_dim)
        attn_dim = 5
        attn = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        
        # theta_vals: (batch, 1, attn_dim*attn_dim)
        theta_vals = torch.randn(batch_size, 1, attn_dim * attn_dim, device=device)
        
        # z_vals: (batch, 8, attn_dim*attn_dim) where first 4 is mu, last 4 is logstd
        z_vals = torch.randn(batch_size, 8, attn_dim * attn_dim, device=device)
        
        sampled_attn = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        
        return attn, sampled_attn, theta_vals, z_vals

# Create encoder for 'attention+offsets' case
class AttentionOffsetsEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = nn.Linear(12, 64)
    
    def forward(self, y, device):
        batch_size = y.shape[0]
        attn_dim = 5
        
        # Return: attn, ?, ?, ?, ?, theta_vals, z_vals
        attn = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        
        theta_vals = torch.randn(batch_size, 1, attn_dim * attn_dim, device=device)
        z_vals = torch.randn(batch_size, 8, attn_dim * attn_dim, device=device)
        
        placeholder1 = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        placeholder2 = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        placeholder3 = torch.randn(batch_size, 1, attn_dim, attn_dim, device=device)
        
        return attn, placeholder1, placeholder2, placeholder3, placeholder3, theta_vals, z_vals

# Create base coordinates (pixel grid)
# x should be (num_pixels, 2)
x = torch.tensor([
    [0.0, 0.0],
    [1.0, 0.0],
    [2.0, 0.0],
    [3.0, 0.0],
    [4.0, 0.0],
    [5.0, 0.0]
], dtype=torch.float32)

# Create input y (batch_size=2, flattened_pixels=12)
y = torch.randn(2, 12, dtype=torch.float32)

# Create encoder model
encoder_model = UnimodalEncoder()

# Device
device = 'cpu'

ARGS = {
    'x': x,
    'y': y,
    'encoder_model': encoder_model,
    't_inf': 'unimodal',
    'r_inf': 'unimodal',
    'device': device,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_mode Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/mrc.py
code served (permissive licence) · get_code("48088186a5bb2896")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# The function get_mode takes a single parameter 'dtype' which should be a numpy dtype
# It checks the dtype against various numpy data types and returns a corresponding mode integer

# We'll create ARGS with a valid dtype that matches one of the conditions
# Let's use np.int8 which returns mode 0

ARGS = {
    'dtype': np.int8
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
load_images Ran
ran on a synthesized input
smlc-nysbc/target-vae/clustering_dsprites.py
code served (permissive licence) · get_code("8c582ad5126b1725")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (2, 4, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np
import tempfile
import os

# Create a temporary .npy file for testing
temp_dir = tempfile.mkdtemp()
temp_npy_path = os.path.join(temp_dir, "test_images.npy")

# Create dummy image data and save it
dummy_images = np.random.rand(2, 4, 4).astype(np.float32)
np.save(temp_npy_path, dummy_images)

ARGS = {
    "path": temp_npy_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 4, 4) — shape (2, 4, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
normalize Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/image.py
code served (permissive licence) · get_code("39e9adc0e88f7f86")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 3, dtype family float, shape (2, 4, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import numpy as np

# Create a small test stack with shape (batch, height, width)
# Using distinct sizes: batch=2, height=4, width=3
stack = np.random.randn(2, 4, 3).astype(np.float32)

# Optionally test with a specific radius, or leave as None to use default
radius = None

ARGS = {
    "stack": stack,
    "radius": radius,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 3, dtype family float, shape (2, 4, 3) — shape (2, 4, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
parse_ctf Ran
ran on a synthesized input
smlc-nysbc/target-vae/src/ctf.py
code served (permissive licence) · get_code("9ef31dc8425cf565")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import io

# Create a CSV string with CTF parameters in whitespace-separated format
csv_content = """1.5 2.7 200.0 1.2 50.0 0.1 0.05 45.0
2.0 2.7 200.0 1.2 60.0 0.15 0.08 50.0"""

# Create a file-like object from the string
f = io.StringIO(csv_content)

ARGS = {
    'f': f
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
ctf_filter Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/src/ctf.py
code served (permissive licence) · get_code("1fd769897b5c1e01")
How it ran

no invocation recorded.

eval_minibatch Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_dsprites.py
code served (permissive licence) · get_code("ca7bf21d29efd052")
How it ran

no invocation recorded.

eval_minibatch Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_galaxy.py
code served (permissive licence) · get_code("f7bf32be6ed3b3ce")
How it ran

no invocation recorded.

eval_model Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_dsprites.py
code served (permissive licence) · get_code("a0c4d1584fb072d5")
How it ran

no invocation recorded.

eval_model Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_galaxy.py
code served (permissive licence) · get_code("9769b6fef90de4d5")
How it ran

no invocation recorded.

get_latent Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/clustering_mnist.py
code served (permissive licence) · get_code("c5f1b20bf55ae6d8")
How it ran

no invocation recorded.

make_header Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/src/mrc.py
code served (permissive licence) · get_code("3e39b8fba4e5dc36")
How it ran

no invocation recorded.

parse Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/src/mrc.py
code served (permissive licence) · get_code("e23c9b574323efb9")
How it ran

no invocation recorded.

train_epoch Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_dsprites.py
code served (permissive licence) · get_code("fef77345332c002c")
How it ran

no invocation recorded.

train_epoch Not yet run
unverified (harvested, no recorded run)
smlc-nysbc/target-vae/train_galaxy.py
code served (permissive licence) · get_code("d4b03d98b21b85d0")
How it ran

no invocation recorded.

Also present in other papers' code

Of 20 samples lifted from this paper's linked repositories, 20 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 17 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

In many imaging modalities, objects of interest can occur in a variety of locations and poses (i.e. are subject to translations and rotations in 2d or 3d), but the location and pose of an object does not change its semantics (i.e. the object's essence). That is, the specific location and rotation of an airplane in satellite imagery, or the 3d rotation of a chair in a natural image, or the rotation of a particle in a cryo-electron micrograph, do not change the intrinsic nature of those objects. Here, we consider the problem of learning semantic representations of objects that are invariant to pose and location in a fully unsupervised manner. We address shortcomings in previous approaches to this problem by introducing TARGET-VAE, a translation and rotation group-equivariant variational autoencoder framework. TARGET-VAE combines three core innovations: 1) a rotation and translation group-equivariant encoder architecture, 2) a structurally disentangled distribution over latent rotation, translation, and a rotation-translation-invariant semantic object representation, which are jointly inferred by the approximate inference network, and 3) a spatially equivariant generator network. In comprehensive experiments, we show that TARGET-VAE learns disentangled representations without supervision that significantly improve upon, and avoid the pathologies of, previous methods. When trained on images highly corrupted by rotation and translation, the semantic representations learned by TARGET-VAE are similar to those learned on consistently posed objects, dramatically improving clustering in the semantic latent space. Furthermore, TARGET-VAE is able to perform remarkably accurate unsupervised pose and location inference. We expect methods like TARGET-VAE will underpin future approaches for unsupervised object generation, pose prediction, and object detection.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2210.12918")
get_code_for_paper("2210.12918")
have("2210.12918")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2210.12918.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2210.12918.svg)](https://syntology.ai/paper/2210.12918)

Connect an agent — have() is free.