SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.02634 · 2021

Light Field Networks: Neural Scene Representations with Single-Evaluation Rendering

arXiv · PDF · Open in the Atlas

Code that ran

11 of 11 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; 1 failure of our instrument; 0 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/2106.02634.json · badge: /api/ran/2106.02634.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
vsitzmann/light-field-networks community (classifier label)
classifier basis not recorded on the link
11 ran · 0 no recorded run
FunctionStatusWhere it lives
class_string_2_class_id Ran
ran on a synthesized input
vsitzmann/light-field-networks/multiclass_dataio.py
code served (permissive licence) · get_code("39d9d9a53aa724b7")
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 torch

string2class_dict = {'02691156': 0, '02958343': 1, '03636649': 2, '04256520': 3, '04530566': 4, '02828884': 5, '03001627': 6, '03691459': 7, '04379243': 8, '02933112': 9, '03211117': 10, '04090263': 11, '04401088': 12}

def class_string_2_class_id(x):
    return string2class_dict[x]

ARGS = {
    'x': '02691156'
}
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 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
closest_to_origin Ran
ran on a synthesized input
vsitzmann/light-field-networks/geometry.py
code served (permissive licence) · get_code("43f00a2b03f12c8a")
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 2, dtype family float, shape (2, 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 torch

torch.manual_seed(0)

# plucker_coord is a tensor with shape (..., 6)
# where the last dimension splits into direction (3,) and moment (3,)
# The function uses torch.cross with dim=-1, which requires the last dimension to be size 3
# We'll create a batch of 2 plucker coordinates

plucker_coord = torch.randn(2, 6)

ARGS = {
    "plucker_coord": plucker_coord
}
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 2, dtype family float, shape (2, 3) — shape (2, 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_num_instances Ran
ran on a synthesized input
vsitzmann/light-field-networks/hdf5_dataio.py
code served (permissive licence) · get_code("b37c86f63a99f99a")
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 tempfile
import os
from glob import glob
import h5py

# Create a temporary directory structure for testing
temp_dir = tempfile.mkdtemp()

# Test case 1: HDF5 file path
hdf5_file = os.path.join(temp_dir, "data.hdf5")
with h5py.File(hdf5_file, 'w') as f:
    f.create_dataset('instance_1', data=[1, 2, 3])
    f.create_dataset('instance_2', data=[4, 5, 6])
    f.create_dataset('instance_3', data=[7, 8, 9])

# Test case 2: Directory structure with subdirectories
data_dir = os.path.join(temp_dir, "data_dir")
os.makedirs(data_dir, exist_ok=True)
os.makedirs(os.path.join(data_dir, "subdir1"), exist_ok=True)
os.makedirs(os.path.join(data_dir, "subdir2"), exist_ok=True)
os.makedirs(os.path.join(data_dir, "subdir3"), exist_ok=True)

# Define ARGS with the hdf5 file path
ARGS = {
    "data_root": hdf5_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
h5py — 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
get_num_instances Ran
ran on a synthesized input
vsitzmann/light-field-networks/multiclass_dataio.py
code served (permissive licence) · get_code("3e1e570476a17777")
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 os
import tempfile
from pathlib import Path

# Create temporary directory structure with necessary files
temp_dir = tempfile.mkdtemp()

# Create object class directories with list files
object_class_1 = os.path.join(temp_dir, "class_1/")
object_class_2 = os.path.join(temp_dir, "class_2/")

os.makedirs(object_class_1, exist_ok=True)
os.makedirs(object_class_2, exist_ok=True)

# Create softras list files for dataset_type='train'
dataset_type = 'train'

# Create list file for class_1
list_file_1 = os.path.join(object_class_1, f'softras_{dataset_type}.lst')
with open(list_file_1, 'w') as f:
    f.write('instance_001\ninstance_002\ninstance_003')

# Create list file for class_2
list_file_2 = os.path.join(object_class_2, f'softras_{dataset_type}.lst')
with open(list_file_2, 'w') as f:
    f.write('instance_004\ninstance_005')

ARGS = {
    'root_dir': temp_dir,
    'dataset_type': 'train'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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
get_ray_origin Ran
ran on a synthesized input
vsitzmann/light-field-networks/geometry.py
code served (permissive licence) · get_code("63cb0313b04ef0e5")
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 2, dtype family float, shape (2, 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 torch

torch.manual_seed(0)

# cam2world is indexed with [..., :3, 3], which suggests it's a tensor
# with at least 2 dimensions where the last two are 4x4 or similar
# We need shape (..., 4, 4) to extract [..., :3, 3]
# Let's create a batch of 2 camera matrices with shape (2, 4, 4)

cam2world = torch.randn(2, 4, 4)

ARGS = {
    "cam2world": cam2world
}
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 2, dtype family float, shape (2, 3) — shape (2, 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
image_loss Ran
ran on a synthesized input
vsitzmann/light-field-networks/loss_functions.py
code served (permissive licence) · get_code("e044b869782c53a7")
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 0, dtype family float, 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 torch
import torch.nn as nn

torch.manual_seed(0)

# Create small tensors for model output and ground truth
batch_size = 2
rgb_dim = 4

# model_out is a dict with 'rgb' key containing a tensor
model_out = {
    'rgb': torch.randn(batch_size, rgb_dim)
}

# gt is a dict with 'rgb' key containing a tensor
gt = {
    'rgb': torch.randn(batch_size, rgb_dim)
}

# mask is optional and not used in the function body, so None is fine
mask = None

ARGS = {
    'model_out': model_out,
    'gt': gt,
    'mask': mask
}
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 0, dtype family float, 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_depth Ran
ran on a synthesized input
vsitzmann/light-field-networks/data_util.py
code served (permissive licence) · get_code("83666af2324d73b9")
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 (1, 32, 32) 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 cv2
import numpy as np
import tempfile
import os

# Create a temporary depth image file
temp_dir = tempfile.mkdtemp()
temp_path = os.path.join(temp_dir, "depth.png")

# Create a simple 16-bit depth image (grayscale)
depth_data = np.random.randint(0, 10000, size=(64, 64), dtype=np.uint16)
cv2.imwrite(temp_path, depth_data)

ARGS = {
    "path": temp_path,
    "sidelength": 32
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, 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 (1, 32, 32) — shape (1, 32, 32) 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_pose Ran
ran on a synthesized input
vsitzmann/light-field-networks/data_util.py
code served (permissive licence) · get_code("313ece65b8317903")
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, 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 file with pose data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "pose.txt")

# Write a single line with 16 space-separated values (4x4 matrix flattened)
pose_values = [str(float(i)) for i in range(16)]
with open(temp_file, 'w') as f:
    f.write(' '.join(pose_values))

ARGS = {
    'filename': temp_file
}
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, 4) — shape (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
load_rgb Ran
ran on a synthesized input
vsitzmann/light-field-networks/data_util.py
code served (permissive licence) · get_code("0952740a9ea73ed3")
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 (100, 100, 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 cv2
import numpy as np
import imageio
import skimage
import tempfile
import os

# Create a temporary RGB image file for testing
temp_dir = tempfile.mkdtemp()
temp_img_path = os.path.join(temp_dir, "test_img.png")

# Create a simple RGB image (height=100, width=120, channels=3)
test_img = np.random.rand(100, 120, 3) * 255
test_img = test_img.astype(np.uint8)
imageio.imwrite(temp_img_path, test_img)

ARGS = {
    "path": temp_img_path,
    "sidelength": None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, imageio, numpy, skimage — 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 (100, 100, 3) — shape (100, 100, 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
normalize_imagenet Ran
ran; our contract draft was wrong, not the code
vsitzmann/light-field-networks/conv_modules.py
pointer only (licence: MIT); the stored copy get_code serves, from tianfr/Semantic-Flow, is pointer only (licence: NOASSERTION) · get_code("943737ea63cae867")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from tianfr/Semantic-Flow, and GitHub could not classify the licence of that repository (NOASSERTION). The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

plucker_embedding Ran
ran on a synthesized input
vsitzmann/light-field-networks/geometry.py
code served (permissive licence) · get_code("1879debc023ea214")
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 3, dtype family float, shape (2, 3, 6) 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
import torch.nn as nn

torch.manual_seed(0)

# Create minimal valid tensors for the function parameters
batch_size = 2
num_points = 3
height = 4
width = 4

# cam2world: (batch_size, 4, 4) - camera to world transformation matrix
cam2world = torch.eye(4).unsqueeze(0).expand(batch_size, 4, 4).clone()
# Add some variation to the first batch element
cam2world[0, :3, 3] = torch.tensor([0.1, 0.2, 0.3])

# intrinsics: (batch_size, 4, 4) - camera intrinsic matrix
# Standard form: [[fx, 0, cx, 0], [0, fy, cy, 0], [0, 0, 1, 0], [0, 0, 0, 1]]
intrinsics = torch.zeros(batch_size, 4, 4)
intrinsics[:, 0, 0] = 500.0  # fx
intrinsics[:, 1, 1] = 500.0  # fy
intrinsics[:, 0, 2] = 320.0  # cx
intrinsics[:, 1, 2] = 240.0  # cy
intrinsics[:, 2, 2] = 1.0
intrinsics[:, 3, 3] = 1.0

# uv: (batch_size, num_points, 2) - pixel coordinates
uv = torch.tensor([
    [[100.0, 150.0], [200.0, 200.0], [300.0, 250.0]],
    [[120.0, 160.0], [210.0, 210.0], [310.0, 260.0]]
], dtype=torch.float32)

ARGS = {
    "cam2world": cam2world,
    "uv": uv,
    "intrinsics": intrinsics,
}
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 3, dtype family float, shape (2, 3, 6) — shape (2, 3, 6) 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

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 1 later) and 8 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.

FunctionAlso present inEarliest-datedOthers named
normalize_imagenet 7 other papers
3 dated earlier, 4 later · 1 only as an assembly brick
1911.05063 (2019-11)
Kaolin: A PyTorch Library for Accelerating 3D Deep Learning Research
2007.10973 (2020-07)
Neural Mesh Flow: 3D Manifold Mesh Generation via Diffeomorphic Flows
2104.09224 (2021-04)
Multi-Modal Fusion Transformer for End-to-End Autonomous Driving
and 4 more not named here
load_depth 1 other paper
1 dated earlier
1906.01618 (2019-06)
Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations
load_pose 1 other paper
1 dated earlier
1906.01618 (2019-06)
Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations

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

Inferring representations of 3D scenes from 2D observations is a fundamental problem of computer graphics, computer vision, and artificial intelligence. Emerging 3D-structured neural scene representations are a promising approach to 3D scene understanding. In this work, we propose a novel neural scene representation, Light Field Networks or LFNs, which represent both geometry and appearance of the underlying 3D scene in a 360-degree, four-dimensional light field parameterized via a neural implicit representation. Rendering a ray from an LFN requires only a single network evaluation, as opposed to hundreds of evaluations per ray for ray-marching or volumetric based renderers in 3D-structured neural scene representations. In the setting of simple scenes, we leverage meta-learning to learn a prior over LFNs that enables multi-view consistent light field reconstruction from as little as a single image observation. This results in dramatic reductions in time and memory complexity, and enables real-time rendering. The cost of storing a 360-degree light field via an LFN is two orders of magnitude lower than conventional methods such as the Lumigraph. Utilizing the analytical differentiability of neural implicit representations and a novel parameterization of light space, we further demonstrate the extraction of sparse depth maps from LFNs.

For agents

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

get_harvested_code_for_paper("2106.02634")
get_code_for_paper("2106.02634")
have("2106.02634")

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

curl https://syntology.ai/api/ran/2106.02634.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.02634.svg)](https://syntology.ai/paper/2106.02634)

Connect an agent — have() is free.