SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1911.08299 · 2019

Learning Modulated Loss for Rotated Object Detection

arXiv · PDF · Open in the Atlas

Code that ran

7 of 13 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 7 have no contract verdict; 0 failures of our instrument; 6 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/1911.08299.json · badge: /api/ran/1911.08299.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
Mrqianduoduo/RSDet-8P-4R community (classifier label)
classifier basis not recorded on the link
7 ran · 6 no recorded run
FunctionStatusWhere it lives
bbox_transform Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/box_utils/bbox_transform.py
code served (permissive licence) · get_code("d2d339a252ba46a7")
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 (2, 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

# Create example bounding boxes in format [x1, y1, x2, y2]
# ex_rois: example/anchor ROIs (batch size 2)
ex_rois = np.array([
    [10.0, 10.0, 50.0, 50.0],
    [20.0, 20.0, 80.0, 80.0]
], dtype=np.float32)

# gt_rois: ground truth ROIs (batch size 2)
gt_rois = np.array([
    [12.0, 12.0, 52.0, 52.0],
    [22.0, 22.0, 82.0, 82.0]
], dtype=np.float32)

# scale_factors: optional scaling for the targets [dx, dy, dw, dh]
scale_factors = [10.0, 10.0, 5.0, 5.0]

ARGS = {
    "ex_rois": ex_rois,
    "gt_rois": gt_rois,
    "scale_factors": scale_factors
}
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 (2, 4) — shape (2, 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
bbox_transform_inv Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/box_utils/bbox_transform.py
code served (permissive licence) · get_code("41aa50153cd73204")
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 EagerTensor, ndim 2, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

# Create small test tensors for bounding boxes and deltas
# boxes: [batch_size, 4] where each row is [xmin, ymin, xmax, ymax]
boxes = tf.constant([
    [10.0, 20.0, 50.0, 80.0],
    [30.0, 40.0, 100.0, 150.0]
], dtype=tf.float32)

# deltas: [batch_size, 4] where each row is [dx, dy, dw, dh]
deltas = tf.constant([
    [0.1, 0.2, 0.3, 0.1],
    [0.05, 0.15, 0.2, 0.25]
], dtype=tf.float32)

# scale_factors: optional scaling factors for deltas
scale_factors = [1.0, 1.0, 1.0, 1.0]

ARGS = {
    "boxes": boxes,
    "deltas": deltas,
    "scale_factors": scale_factors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
enum_ratios Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/box_utils/anchor_utils.py
code served (permissive licence) · get_code("07832d24c2b725e2")
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 3, 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 tensorflow as tf
import numpy as np

# Set seed for reproducibility
tf.random.set_seed(0)

# Create small anchor tensors
# anchors should be a tensor with shape (N, 4) where columns are indexed
# The function accesses anchors[:, 2] (ws) and anchors[:, 3] (hs)
anchors = tf.constant([
    [0.0, 0.0, 10.0, 20.0],
    [0.0, 0.0, 15.0, 30.0]
], dtype=tf.float32)

# anchor_ratios should be a list or array of ratio values
anchor_ratios = [0.5, 1.0, 2.0]

ARGS = {
    "anchors": anchors,
    "anchor_ratios": anchor_ratios
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 3, 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
enum_scales Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/box_utils/anchor_utils.py
code served (permissive licence) · get_code("d7d04d0980025fa0")
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 SymbolicTensor. 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 tensorflow as tf
import numpy as np

# Create base_anchor as a TensorFlow tensor
# The function multiplies it with anchor_scales, so it should be compatible
# Looking at the function: base_anchor * tf.constant(..., shape=(len(anchor_scales), 1))
# This suggests base_anchor should be broadcastable with (n, 1)
# A typical anchor would be shape (4,) for [x, y, w, h]
base_anchor = tf.constant([1.0, 2.0, 4.0, 8.0], dtype=tf.float32)

# anchor_scales is a list/array of scale values
# The function converts it to a constant with shape (len(anchor_scales), 1)
anchor_scales = [0.5, 1.0, 2.0]

ARGS = {
    "base_anchor": base_anchor,
    "anchor_scales": anchor_scales,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor 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
focal_loss Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/losses/losses.py
code served (permissive licence) · get_code("e57f95591345ae49")
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 SymbolicTensor. 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 tensorflow as tf
import numpy as np

# Create small test tensors
batch_size = 2
num_classes = 4

# labels: binary labels for each prediction (shape: [batch_size, num_classes])
labels = tf.constant([
    [1.0, 0.0, 1.0, 0.0],
    [0.0, 1.0, 0.0, 1.0]
], dtype=tf.float32)

# pred: logits/predictions (shape: [batch_size, num_classes])
pred = tf.constant([
    [2.1, -0.5, 1.3, -1.2],
    [-0.8, 1.5, 0.2, 2.0]
], dtype=tf.float32)

# anchor_state: state indicator (-1: ignore, 0: negative, 1: positive)
# Must have at least one element equal to 1 for normalizer to work
anchor_state = tf.constant([1, 1, 0, -1], dtype=tf.int32)

ARGS = {
    "labels": labels,
    "pred": pred,
    "anchor_state": anchor_state,
    "alpha": 0.25,
    "gamma": 2.0
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type SymbolicTensor — a SymbolicTensor 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
focal_loss_ Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/losses/losses.py
code served (permissive licence) · get_code("33ffc6daf73c0d55")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Set seeds for reproducibility
tf.random.set_seed(0)
np.random.seed(0)

# Create small tensors for testing
batch_size = 2
num_anchors = 4

# labels: binary labels for each anchor (0 or 1)
labels = tf.constant([1, 0, 1, 0], dtype=tf.int32)

# pred: logits/predictions for each anchor
pred = tf.constant([0.5, -0.3, 2.1, -1.2], dtype=tf.float32)

# anchor_state: state of anchors (-1 means ignore, otherwise valid)
# Using -1 to mark some anchors to be filtered out, and other values for valid anchors
anchor_state = tf.constant([-1, 1, 1, 0], dtype=tf.int32)

# alpha and gamma are optional parameters with defaults
alpha = 0.25
gamma = 2.0

ARGS = {
    "labels": labels,
    "pred": pred,
    "anchor_state": anchor_state,
    "alpha": alpha,
    "gamma": gamma,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
rbbox_transform_inv Ran
ran on a synthesized input
Mrqianduoduo/RSDet-8P-4R/libs/box_utils/bbox_transform.py
code served (permissive licence) · get_code("33ffa5657fe69075")
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 EagerTensor, ndim 2, dtype family float, shape (2, 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 tensorflow as tf
import numpy as np

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

# Create small test tensors for boxes and deltas
# boxes shape: (batch_size, 5) - [ctr_x, ctr_y, w, h, theta]
boxes = tf.constant([
    [10.0, 20.0, 5.0, 8.0, 0.5],
    [30.0, 40.0, 6.0, 9.0, 1.2]
], dtype=tf.float32)

# deltas shape: (batch_size, 5) - [dx, dy, dw, dh, dtheta]
deltas = tf.constant([
    [0.1, 0.2, 0.3, 0.15, 0.05],
    [0.05, 0.1, 0.2, 0.25, 0.1]
], dtype=tf.float32)

# scale_factors - optional parameter, provide as a list
scale_factors = [1.0, 1.0, 1.0, 1.0, 1.0]

ARGS = {
    "boxes": boxes,
    "deltas": deltas,
    "scale_factors": scale_factors
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (2, 5) — shape (2, 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
XceptionModel Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/networks/xception.bak.py
code served (permissive licence) · get_code("cd5c59c480f8b497")
How it ran

no invocation recorded.

norm Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/networks/opts.py
code served (permissive licence) · get_code("a4a6ff951a014092")
How it ran

no invocation recorded.

reduced_kernel_size_for_small_input Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/networks/xception.bak.py
code served (permissive licence) · get_code("1d0c0051ecdfaf14")
How it ran

no invocation recorded.

relu_separable_bn_block Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/networks/xception.bak.py
code served (permissive licence) · get_code("23789094dc8fdfdd")
How it ran

no invocation recorded.

relu_separable_bn_block Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/networks/xception.py
code served (permissive licence) · get_code("46be84834b582dd8")
How it ran

no invocation recorded.

smooth_l1_loss_rcnn Not yet run
unverified (harvested, no recorded run)
Mrqianduoduo/RSDet-8P-4R/libs/losses/losses.py
code served (permissive licence) · get_code("9ce96a058bfde15b")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 0 the same month, 0 later) and 11 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
enum_scales 1 other paper
1 dated earlier
1612.03144 (2016-12)
Feature Pyramid Networks for Object Detection
reduced_kernel_size_for_small_input 1 other paper
1 dated earlier
1610.02357 (2016-10)
Xception: Deep Learning with Depthwise Separable Convolutions

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

Popular rotated detection methods usually use five parameters (coordinates of the central point, width, height, and rotation angle) to describe the rotated bounding box and l1-loss as the loss function. In this paper, we argue that the aforementioned integration can cause training instability and performance degeneration, due to the loss discontinuity resulted from the inherent periodicity of angles and the associated sudden exchange of width and height. This problem is further pronounced given the regression inconsistency among five parameters with different measurement units. We refer to the above issues as rotation sensitivity error (RSE) and propose a modulated rotation loss to dismiss the loss discontinuity. Our new loss is combined with the eight-parameter regression to further solve the problem of inconsistent parameter regression. Experiments show the state-of-art performances of our method on the public aerial image benchmark DOTA and UCAS-AOD. Its generalization abilities are also verified on ICDAR2015, HRSC2016, and FDDB. Qualitative improvements can be seen in Fig 1, and the source code will be released with the publication of the paper.

For agents

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

get_harvested_code_for_paper("1911.08299")
get_code_for_paper("1911.08299")
have("1911.08299")

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

curl https://syntology.ai/api/ran/1911.08299.json

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

[![Syntology run record](https://syntology.ai/api/ran/1911.08299.svg)](https://syntology.ai/paper/1911.08299)

Connect an agent — have() is free.