SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1904.01786 · 2019

Soft Rasterizer: A Differentiable Renderer for Image-based 3D Reasoning

arXiv · PDF · Open in the Atlas

Code that ran

9 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 8 have no contract verdict; 1 failure of our instrument; 2 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/1904.01786.json · badge: /api/ran/1904.01786.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
ShichenLiu/SoftRas canonical (classifier label)
classifier basis: census
8 ran · 1 no recorded run
kach/softraxterizer community (classifier label)
classifier basis not recorded on the link
1 ran · 1 no recorded run
FunctionStatusWhere it lives
directional_lighting Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/functional/directional_lighting.py
code served (permissive licence) · get_code("74684204a6a3b4c6")
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, 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 torch
import torch.nn.functional as F
import numpy as np

torch.manual_seed(0)

# Create small tensors for the function
# light: shape (batch_size, num_points, 3) - RGB values
batch_size = 2
num_points = 4
light = torch.randn(batch_size, num_points, 3, dtype=torch.float32)

# normals: shape (batch_size, num_points, 3) - normal vectors
normals = torch.randn(batch_size, num_points, 3, dtype=torch.float32)
normals = F.normalize(normals, dim=2)  # normalize to unit vectors

# light_intensity: scalar
light_intensity = 0.5

# light_color: tuple or list or tensor (will be converted to tensor)
light_color = (1.0, 1.0, 1.0)

# light_direction: tuple or list or tensor (will be converted to tensor)
light_direction = (0.0, 1.0, 0.0)

ARGS = {
    "light": light,
    "normals": normals,
    "light_intensity": light_intensity,
    "light_color": light_color,
    "light_direction": light_direction,
}
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 Tensor, 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
face_vertices Ran
ran; our contract draft was wrong, not the code
ShichenLiu/SoftRas/soft_renderer/functional/face_vertices.py
pointer only (licence: MIT); the stored copy get_code serves, from ShenhanQian/GaussianAvatars, is pointer only (licence: NOASSERTION) · get_code("42d983df1501fbb5")
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 ShenhanQian/GaussianAvatars, 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.

get_pixel Ran
ran on a synthesized input
kach/softraxterizer/softraxterizer.py
code served (permissive licence) · get_code("01749edd95bc6e58")
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 jax.numpy as np

ARGS = {
    "left": -1.0,
    "top": 1.0,
    "right": 1.5,
    "bottom": -1.5,
    "xres": 50,
    "yres": 50,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
jax — 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_points_from_angles Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/functional/get_points_from_angles.py
code served (permissive licence) · get_code("8497e22d3575f176")
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 math
import torch

torch.manual_seed(0)

# Test case 1: scalar distance with degrees=True
distance_scalar = 10.0
elevation_scalar = 45.0
azimuth_scalar = 30.0
degrees_scalar = True

# Test case 2: tensor distance with degrees=True
distance_tensor = torch.tensor([5.0, 8.0])
elevation_tensor = torch.tensor([30.0, 60.0])
azimuth_tensor = torch.tensor([45.0, 90.0])
degrees_tensor = True

# Use the tensor version for ARGS
ARGS = {
    "distance": distance_tensor,
    "elevation": elevation_tensor,
    "azimuth": azimuth_tensor,
    "degrees": degrees_tensor
}
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
load_mtl Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/functional/load_obj.py
code served (permissive licence) · get_code("158489df7e42d7e0")
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, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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

# Create a temporary MTL file for testing
temp_dir = tempfile.mkdtemp()
mtl_filename = os.path.join(temp_dir, "test.mtl")

mtl_content = """newmtl material1
map_Kd texture1.png
Kd 0.8 0.6 0.4

newmtl material2
map_Kd texture2.jpg
Kd 0.1 0.2 0.3

newmtl material3
Kd 0.5 0.5 0.5
"""

with open(mtl_filename, 'w') as f:
    f.write(mtl_content)

ARGS = {
    'filename_mtl': mtl_filename
}
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 tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
look Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/functional/look.py
code served (permissive licence) · get_code("e13da960399c788d")
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, 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 torch
import numpy as np

torch.manual_seed(0)

# Create valid arguments for the look function
vertices = torch.randn(2, 4, 3, dtype=torch.float32)  # batch=2, num_vertices=4, dim=3

eye = torch.tensor([[0.0, 0.0, 5.0], [1.0, 1.0, 5.0]], dtype=torch.float32)  # batch=2, dim=3

direction = torch.tensor([[0.0, 1.0, 0.0], [0.0, 1.0, 0.0]], dtype=torch.float32)  # batch=2, dim=3

up = torch.tensor([[0.0, 0.0, 1.0], [0.0, 0.0, 1.0]], dtype=torch.float32)  # batch=2, dim=3

ARGS = {
    "vertices": vertices,
    "eye": eye,
    "direction": direction,
    "up": up
}
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 Tensor, 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
look_at Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/functional/look_at.py
code served (permissive licence) · get_code("d7e7c5fad58018a4")
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, 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 torch
import numpy as np

torch.manual_seed(0)

# Create vertices tensor with shape (batch_size, num_vertices, 3)
# batch_size=2, num_vertices=4, coordinates=3
vertices = torch.randn(2, 4, 3, dtype=torch.float32)

# Create eye parameter - can be list, tuple, numpy array, or tensor
eye = [1.0, 2.0, 3.0]

# Create at parameter (look-at point) - defaults to origin
at = [0.0, 0.0, 0.0]

# Create up parameter (up vector) - defaults to y-axis
up = [0.0, 1.0, 0.0]

ARGS = {
    "vertices": vertices,
    "eye": eye,
    "at": at,
    "up": up,
}
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 Tensor, 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
orthogonal Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/transform.py
code served (permissive licence) · get_code("e26d5f3fd5f9dd7c")
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, 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 torch

torch.manual_seed(0)

# Create a 3D vertices tensor with shape (batch_size, num_vertices, 3)
# batch_size=2, num_vertices=4, coordinates=3 (x, y, z)
vertices = torch.randn(2, 4, 3)

# Create scale parameter
scale = 2.0

ARGS = {
    "vertices": vertices,
    "scale": scale,
}
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, 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
perspective Ran
ran on a synthesized input
ShichenLiu/SoftRas/soft_renderer/transform.py
code served (permissive licence) · get_code("be8ae6f86624b34a")
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, 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 torch
import math

torch.manual_seed(0)

# The function expects:
# - vertices: a 3D tensor with shape (batch_size, num_vertices, 3)
#   where the last dimension contains [x, y, z] coordinates
# - angle: a float (default 30.0)

# Create a small batch of vertices
# Shape: (batch_size=2, num_vertices=4, coordinates=3)
vertices = torch.randn(2, 4, 3, dtype=torch.float32)

# Ensure z coordinates are non-zero to avoid division by zero
vertices[:, :, 2] = torch.abs(vertices[:, :, 2]) + 0.1

angle = 30.0

ARGS = {
    "vertices": vertices,
    "angle": angle,
}
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, 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
ambient_lighting Not yet run
unverified (harvested, no recorded run)
ShichenLiu/SoftRas/soft_renderer/functional/ambient_lighting.py
code served (permissive licence) · get_code("7ffb516ee103321a")
How it ran

no invocation recorded.

softras Not yet run
unverified (harvested, no recorded run)
kach/softraxterizer/softraxterizer.py
code served (permissive licence) · get_code("b8b5518bfb5539c8")
How it ran

no invocation recorded.

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, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 10 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
face_vertices 3 other papers
3 later
2312.02069 (2023-12)
GaussianAvatars: Photorealistic Head Avatars with Rigged 3D Gaussians
2401.00979 (2024-01)
3D Visibility-Aware Generalizable Neural Radiance Fields for Interacting Hands
2410.08840 (2024-10)
Learning Interaction-aware 3D Gaussian Splatting for One-shot Hand Avatars

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

Rendering bridges the gap between 2D vision and 3D scenes by simulating the physical process of image formation. By inverting such renderer, one can think of a learning approach to infer 3D information from 2D images. However, standard graphics renderers involve a fundamental discretization step called rasterization, which prevents the rendering process to be differentiable, hence able to be learned. Unlike the state-of-the-art differentiable renderers, which only approximate the rendering gradient in the back propagation, we propose a truly differentiable rendering framework that is able to (1) directly render colorized mesh using differentiable functions and (2) back-propagate efficient supervision signals to mesh vertices and their attributes from various forms of image representations, including silhouette, shading and color images. The key to our framework is a novel formulation that views rendering as an aggregation function that fuses the probabilistic contributions of all mesh triangles with respect to the rendered pixels. Such formulation enables our framework to flow gradients to the occluded and far-range vertices, which cannot be achieved by the previous state-of-the-arts. We show that by using the proposed renderer, one can achieve significant improvement in 3D unsupervised single-view reconstruction both qualitatively and quantitatively. Experiments also demonstrate that our approach is able to handle the challenging tasks in image-based shape fitting, which remain nontrivial to existing differentiable renderers.

For agents

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

get_harvested_code_for_paper("1904.01786")
get_code_for_paper("1904.01786")
have("1904.01786")

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

curl https://syntology.ai/api/ran/1904.01786.json

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

[![Syntology run record](https://syntology.ai/api/ran/1904.01786.svg)](https://syntology.ai/paper/1904.01786)

Connect an agent — have() is free.