SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2002.12319 · ICLR · 2020

Semantically-Guided Representation Learning for Self-Supervised Monocular Depth

Jie Li, Vitor Guizilini, Rui Hou, Rares, Ambrus, Adrien Gaidon

arXiv · PDF · Open in the Atlas

Code that ran

14 of 14 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; 6 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 8 failures of our instrument; 0 no recorded run), measured 2026-11-21; 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/2002.12319.json · badge: /api/ran/2002.12319.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
TRI-ML/packnet-sfm canonical (classifier label)
classifier basis not recorded on the link
14 ran · 0 no recorded run
FunctionStatusWhere it lives
construct_K Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/geometry/camera_utils.py
code served (permissive licence) · get_code("3f9936537c112aa9")
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 (3, 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)

ARGS = {
    "fx": 500.0,
    "fy": 500.0,
    "cx": 320.0,
    "cy": 240.0,
    "dtype": torch.float,
    "device": None,
}
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 (3, 3) — shape (3, 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
dummy_calibration Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/datasets/image_dataset.py
code served (permissive licence) · get_code("56344348062c1db9")
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 (3, 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
from PIL import Image

# Create a small dummy image with size method
class DummyImage:
    def __init__(self, width, height):
        self.width = width
        self.height = height
    
    def size(self):
        return (self.width, self.height)

# Create a PIL-like object with a size property
image = Image.new('RGB', (640, 480))

ARGS = {
    'image': image
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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 (3, 3) — shape (3, 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
euler2mat Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/geometry/pose_utils.py
code served (permissive licence) · get_code("a4c5f47d2c6efb9c")
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, 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)

# The function expects a tensor of shape (B, 3) containing euler angles
# where B is the batch size
angle = torch.randn(2, 3)  # batch_size=2, 3 euler angles (x, y, z)

ARGS = {
    "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, 3, 3) — shape (2, 3, 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_idx Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/datasets/image_dataset.py
code served (permissive licence) · get_code("6d3ca8f94497d761")
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 re

def get_idx(filename):
    return int(re.search('\\d+', filename).group())

# The function takes a filename string and extracts the first sequence of digits
# It uses re.search to find digits and .group() to get the match
ARGS = {
    'filename': 'file_123_data.txt'
}
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
invert_pose Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/geometry/pose_utils.py
code served (permissive licence) · get_code("1ec4bb6cee548217")
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, 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)

# Create a batch of 2 pose matrices [B, 4, 4]
# These are valid transformation matrices (rotation + translation)
B = 2
T = torch.eye(4, dtype=torch.float32).unsqueeze(0).repeat(B, 1, 1)

# Add some rotation (identity rotation for simplicity)
# Add some translation
T[0, :3, -1] = torch.tensor([1.0, 2.0, 3.0])
T[1, :3, -1] = torch.tensor([4.0, 5.0, 6.0])

# Verify shape is [B, 4, 4]
assert T.shape == (2, 4, 4)

ARGS = {
    "T": T
}
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, 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
pose_vec2mat Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/geometry/pose_utils.py
code served (permissive licence) · get_code("e077e60a96f5a927")
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, 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 pose_vec2mat expects:
# - vec: a tensor of shape (batch_size, 6) containing [trans_x, trans_y, trans_z, rot_x, rot_y, rot_z]
# - mode: a string specifying rotation mode, default is 'euler'

# Create a small batch of pose vectors
# Shape: (batch_size=2, 6) where first 3 are translation, last 3 are euler angles
vec = torch.randn(2, 6)

# mode parameter - can be 'euler' or None
mode = 'euler'

ARGS = {
    'vec': vec,
    'mode': mode
}
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, 4) — shape (2, 3, 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
read_files Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/datasets/image_dataset.py
code served (permissive licence) · get_code("bdf83cb105e62b4d")
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 defaultdict. 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 os
import tempfile
from collections import defaultdict

# Create a temporary directory structure with some test files
temp_dir = tempfile.mkdtemp()

# Create subdirectories and files
os.makedirs(os.path.join(temp_dir, "subdir1"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "subdir2"), exist_ok=True)
os.makedirs(os.path.join(temp_dir, "empty_subdir"), exist_ok=True)

# Create some image files
open(os.path.join(temp_dir, "image1.png"), 'w').close()
open(os.path.join(temp_dir, "image2.jpg"), 'w').close()
open(os.path.join(temp_dir, "subdir1", "image3.jpeg"), 'w').close()
open(os.path.join(temp_dir, "subdir1", "document.txt"), 'w').close()
open(os.path.join(temp_dir, "subdir2", "image4.PNG"), 'w').close()

ARGS = {
    "directory": temp_dir,
    "ext": ('.png', '.jpg', '.jpeg'),
    "skip_empty": True
}
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 defaultdict — a defaultdict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a defaultdict output the recorded ndim 0 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
read_npz_depth Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/datasets/kitti_dataset.py
code served (permissive licence) · get_code("1c9d1d852e11deaf")
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 (4, 3, 1) 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 .npz file with depth data
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_depth.npz")

# Create sample depth data
depth_data = np.random.rand(4, 3).astype(np.float32)
np.savez(temp_file, stereo_depth=depth_data)

# Define arguments
ARGS = {
    "file": temp_file,
    "depth_type": "stereo"
}
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 (4, 3, 1) — shape (4, 3, 1) 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
resize_depth Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
TRI-ML/packnet-sfm/packnet_sfm/datasets/augmentations.py
code served (permissive licence) · get_code("837c6a474dd26eff")
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 (3, 6, 1) 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

# Create a small depth map (height=5, width=4)
depth_map = np.random.rand(5, 4).astype(np.float32) * 10

# Define target shape (H=3, W=6)
target_shape = (3, 6)

ARGS = {
    'depth': depth_map,
    'shape': target_shape,
}
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 (3, 6, 1) — shape (3, 6, 1) 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
rotx Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/datasets/kitti_dataset_utils.py
code served (permissive licence) · get_code("5c94a7e309621404")
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 (3, 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

# The function rotx takes a single parameter t (float - theta angle)
# and returns a 3x3 rotation matrix about the x-axis.
# The function uses np.cos, np.sin, and np.array operations on t.

ARGS = {
    "t": 0.5  # A simple float angle in radians
}
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 (3, 3) — shape (3, 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
roty Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/datasets/kitti_dataset_utils.py
code served (permissive licence) · get_code("0a19721bdaffaddd")
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 (3, 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

ARGS = {
    "t": 0.5
}
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 (3, 3) — shape (3, 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
rotz Ran
ran, honoured the contract we drafted
TRI-ML/packnet-sfm/packnet_sfm/datasets/kitti_dataset_utils.py
code served (permissive licence) · get_code("e1451f5589db8939")
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 (3, 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

ARGS = {
    "t": 0.5
}
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 (3, 3) — shape (3, 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
scale_intrinsics Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
TRI-ML/packnet-sfm/packnet_sfm/geometry/camera_utils.py
code served (permissive licence) · get_code("8f01cf597a6c5057")
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, 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)

# K is a camera intrinsics matrix
# The function uses advanced indexing: K[..., 0, 0], K[..., 1, 1], etc.
# This suggests K can be batched or single matrix
# Shape should be (..., 3, 3) for intrinsic matrices
# Let's create a batch of 2 intrinsic matrices of shape (2, 3, 3)
K = torch.tensor([
    [[500.0, 0.0, 320.0],
     [0.0, 500.0, 240.0],
     [0.0, 0.0, 1.0]],
    [[600.0, 0.0, 400.0],
     [0.0, 600.0, 300.0],
     [0.0, 0.0, 1.0]]
], dtype=torch.float32)

# x_scale: scalar factor for horizontal scaling
x_scale = 0.5

# y_scale: scalar factor for vertical scaling
y_scale = 0.75

ARGS = {
    'K': K,
    'x_scale': x_scale,
    'y_scale': y_scale
}
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, 3, 3) — shape (2, 3, 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
view_synthesis Ran
ran; our contract draft was wrong, not the code
TRI-ML/packnet-sfm/packnet_sfm/geometry/camera_utils.py
code served (permissive licence) · get_code("8c84d49a14bcb232")
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, 3, 4, 5) 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)

class Camera:
    """Stand-in Camera class with reconstruct and project methods."""
    
    def __init__(self, intrinsics=None):
        self.intrinsics = intrinsics or torch.eye(3)
    
    def reconstruct(self, depth, frame='w'):
        """
        Reconstruct 3D world points from depth map.
        
        Parameters:
        - depth: [B, 1, H, W]
        - frame: str (not used in this stub)
        
        Returns:
        - world_points: [B, 3, H, W] (or similar shape for 3D coordinates)
        """
        B, _, H, W = depth.shape
        # Return dummy world points with shape [B, 3, H, W]
        world_points = torch.randn(B, 3, H, W)
        return world_points
    
    def project(self, world_points, frame='w'):
        """
        Project 3D world points to 2D image coordinates.
        
        Parameters:
        - world_points: [B, 3, H, W]
        - frame: str (not used in this stub)
        
        Returns:
        - coords: [B, H, W, 2] normalized coordinates for grid_sample
        """
        B, _, H, W = world_points.shape
        # Return dummy normalized coordinates in [-1, 1] range for grid_sample
        coords = torch.randn(B, H, W, 2) * 0.5
        return coords

# Create test tensors
B, C, H, W = 2, 3, 4, 5
ref_image = torch.randn(B, C, H, W)
depth = torch.randn(B, 1, H, W)

# Create camera instances
ref_cam = Camera()
cam = Camera()

ARGS = {
    'ref_image': ref_image,
    'depth': depth,
    'ref_cam': ref_cam,
    'cam': cam,
    'mode': 'bilinear',
    'padding_mode': 'zeros'
}
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, 3, 4, 5) — shape (2, 3, 4, 5) 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 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 14 are also present in other papers whose code we hold (5 with at least one paper dated earlier, 3 the same month, 14 later) and 0 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
roty 13 other papers
3 dated earlier, 1 the same month, 9 later
1612.07695 (2016-12)
MultiNet: Real-time Joint Semantic Reasoning for Autonomous Driving
1811.07782 (2018-11)
Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN
1912.12147 (2019-12)
Cooperative Perception for 3D Object Detection in Driving Scenarios using Infrastructure Sensors
and 10 more not named here
rotx 12 other papers
3 dated earlier, 1 the same month, 8 later
1612.07695 (2016-12)
MultiNet: Real-time Joint Semantic Reasoning for Autonomous Driving
1811.07782 (2018-11)
Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN
1912.12147 (2019-12)
Cooperative Perception for 3D Object Detection in Driving Scenarios using Infrastructure Sensors
and 9 more not named here
rotz 8 other papers
3 dated earlier, 1 the same month, 4 later
1612.07695 (2016-12)
MultiNet: Real-time Joint Semantic Reasoning for Autonomous Driving
1811.07782 (2018-11)
Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN
1912.12147 (2019-12)
Cooperative Perception for 3D Object Detection in Driving Scenarios using Infrastructure Sensors
and 5 more not named here
construct_K 2 other papers
1 dated earlier, 1 later
1912.03426 (2019-12)
Self-Supervised 3D Keypoint Learning for Ego-motion Estimation
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
read_npz_depth 2 other papers
2 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
2202.03131 (2022-02)
Transformers in Self-Supervised Monocular Depth Estimation with Unknown Camera Intrinsics
resize_depth 2 other papers
2 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
2202.03131 (2022-02)
Transformers in Self-Supervised Monocular Depth Estimation with Unknown Camera Intrinsics
scale_intrinsics 2 other papers
1 dated earlier, 1 later
1912.03426 (2019-12)
Self-Supervised 3D Keypoint Learning for Ego-motion Estimation
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
dummy_calibration 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
euler2mat 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
get_idx 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
invert_pose 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
pose_vec2mat 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
read_files 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation
view_synthesis 1 other paper
1 later
2103.16694 (2021-03)
Geometric Unsupervised Domain Adaptation for Semantic Segmentation

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 2026-11-21.

Repositories linked to this paper

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

Abstract

Self-supervised learning is showing great promise for monocular depth estimation, using geometry as the only source of supervision. Depth networks are indeed capable of learning representations that relate visual appearance to 3D properties by implicitly leveraging category-level patterns. In this work we investigate how to leverage more directly this semantic structure to guide geometric representation learning, while remaining in the self-supervised regime. Instead of using semantic labels and proxy losses in a multi-task approach, we propose a new architecture leveraging fixed pretrained semantic segmentation networks to guide self-supervised representation learning via pixel-adaptive convolutions. Furthermore, we propose a two-stage training process to overcome a common semantic bias on dynamic objects via resampling. Our method improves upon the state of the art for self-supervised monocular depth prediction over all pixels, fine-grained details, and per semantic categories. †

For agents

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

get_harvested_code_for_paper("2002.12319")
get_code_for_paper("2002.12319")
have("2002.12319")

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

curl https://syntology.ai/api/ran/2002.12319.json

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

[![Syntology run record](https://syntology.ai/api/ran/2002.12319.svg)](https://syntology.ai/paper/2002.12319)

Connect an agent — have() is free.