SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2405.01217 · ICLR · 2024

CromSS: Cross-modal pretraining with noisy labels for remote sensing image segmentation

Xiao Zhu, Yi Wang, Chenying Liu, Conrad Albrecht

arXiv · PDF · Open in the Atlas

Code that ran

5 of 6 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 0 have no contract verdict; 5 failures of our instrument; 1 no recorded run), measured 2026-11-22; 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/2405.01217.json · badge: /api/ran/2405.01217.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
zhu-xlab/CromSS canonical (classifier label)
classifier basis: census
5 ran · 1 no recorded run
FunctionStatusWhere it lives
get_and_modify_module_state_dict_keys Ran
ran; our contract draft was wrong, not the code
zhu-xlab/CromSS/utils/finetune.py
code served (permissive licence) · get_code("fe6c51e29b6c1f1d")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Build a minimal checkpoint dict for 'mocoq' prt_tag
checkpoint_mocoq = {
    'state_dict': {
        'module.encoder_q.layer1.weight': torch.randn(4, 3),
        'module.encoder_q.layer2.bias': torch.randn(4),
        'module.encoder_q.fc.weight': torch.randn(2, 4),
    }
}

# Build a minimal checkpoint dict for 'mfs1' prt_tag with decoder and seghead
checkpoint_mfs1 = {
    'state_dict': {
        'net.encoder1.layer1.weight': torch.randn(4, 3),
        'net.encoder1.layer2.bias': torch.randn(4),
        'net.decoder.layer1.weight': torch.randn(3, 4),
        'net.decoder.layer2.bias': torch.randn(3),
        'net.segmentation_head.layer1.weight': torch.randn(2, 3),
        'net.segmentation_head.layer2.bias': torch.randn(2),
    }
}

# Build a minimal checkpoint dict for 'scalemae' prt_tag
checkpoint_scalemae = {
    'model': {
        'blocks.0.weight': torch.randn(4, 3),
        'cls.token': torch.randn(1, 4),
        'pos.embed': torch.randn(197, 4),
        'patch.embed.weight': torch.randn(4, 3),
        'channel.norm.weight': torch.randn(4),
    }
}

ARGS = {
    'checkpoint': checkpoint_mocoq,
    'prt_tag': 'mocoq',
    'load_decoder': False,
    'load_seghead': False,
    'n_load_decoder_blocks': 5,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
get_device_info Ran
ran; our contract draft was wrong, not the code
zhu-xlab/CromSS/py_scripts_SSL4EO/get_args_finetune.py
code served (permissive licence) · get_code("a37cc1045556a307")
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 Args. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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 torch

class Args:
    def __init__(self):
        self.slurm = False
        self.accelerator = 'cpu'
        self.nodes = None
        self.devices = None
        self.strategy = None
        self.current_device = None

ARGS = {'args': Args()}
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 Args — a Args output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Args 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
modify_module_state_dict_keys Ran
ran; our contract draft was wrong, not the code
zhu-xlab/CromSS/models/pl_pretrain.py
code served (permissive licence) · get_code("3e175d36d9972f8a")
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) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a state dict with some keys that have the 'module.' prefix
# and some that don't, to test the function's behavior
msd = {
    'module.layer1.weight': torch.randn(4, 3),
    'module.layer1.bias': torch.randn(4),
    'module.layer2.weight': torch.randn(3, 2),
    'layer3.weight': torch.randn(2, 4),
    'layer3.bias': torch.randn(2),
}

ARGS = {
    'msd': msd
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
modify_msd_keys Ran
ran; our contract draft was wrong, not the code
zhu-xlab/CromSS/utils/finetune.py
code served (permissive licence) · get_code("331deb960d8cb8de")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Construct a model state dict (msd) - a dictionary mapping parameter names to tensors
msd = {
    "module.fc.weight": torch.randn(4, 3),
    "module.fc.bias": torch.randn(4),
    "module.layer1.weight": torch.randn(3, 2),
    "module.layer1.bias": torch.randn(3),
    "other.weight": torch.randn(2, 2),
    "prefix_fc.weight": torch.randn(5, 4),
    "prefix_other.weight": torch.randn(3, 3),
}

# The function modifies keys in the msd dict
# It looks for keys starting with prefix, and replaces prefix with replace
# Special condition: it doesn't keep keys that start with '{replace}fc'
prefix = "module."
replace = "backbone."

ARGS = {
    "msd": msd,
    "prefix": prefix,
    "replace": replace,
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
set_log_files Ran
ran; our contract draft was wrong, not the code
zhu-xlab/CromSS/py_scripts_SSL4EO/get_args_finetune.py
code served (permissive licence) · get_code("14eeb092dcc5e411")
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 Args. 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 sys
import json
import numpy as np
from pathlib import Path

class Args:
    def __init__(self):
        self.dataset = 'DFC'
        self.n_classes = 8
        self.lbl_t = 'dfc'
        self.train_path = 'path_s_100_train'
        self.mt = 'resnet50'
        self.n_channels = 3
        self.prt_tag = 'rand'
        self.frm = 'unet'
        self.patch_size = 16
        self.ws2b = 32
        self.wpath = 'weights_path'
        self.wls = False
        self.wlsf = [0.1]
        self.wls_prior_type = 'none'
        self.wss = False
        self.wss_prop = 0.8
        self.wss_ctype = 'random'
        self.wss_rm_func = 'none'
        self.wss_nep = 5
        self.wschdt = 'none'
        self.wschdv = 0.1
        self.wbs = 16
        self.wep = 10
        self.wload_tag = 'val'
        self.lr = 0.001
        self.eft_tag = 'noeft'
        self.eft_ep = 0
        self.dft_tag = 'nodft'
        self.epochs = 50
        self.job_type = 'default'
        self.schdt = 'none'
        self.schdv = [10, 20]
        self.it = 's1'
        self.s1_norm = False
        self.s2_norm = False
        self.only_test = False
        self.ts_wpath = None
        self.save_dir = '/tmp/test_save'
        self.pdir = '/tmp/test_pdir'
        self.mcr = 0.5
        self.seed = 42
        self.bs = 16
        self.unet_type = 'default_unet'

ARGS = {
    'args': Args()
}
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 Args — a Args output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Args 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
get_model_state_dict_from_saved_models Not yet run
unverified (harvested, no recorded run)
zhu-xlab/CromSS/utils/finetune.py
code served (permissive licence) · get_code("8c3e8290362d42fc")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 5 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
modify_module_state_dict_keys 1 other paper
1 dated earlier
2403.01641 (2024-03)
AIO2: Online Correction of Object Labels for Deep Learning with Incomplete Annotation in Remote Sensing Image 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 2026-11-22.

Repositories linked to this paper

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

Abstract

We explore the potential of large-scale noisily labeled data to enhance feature learning by pretraining semantic segmentation models within a multi-modal framework for geospatial applications. We propose a novel Cross-modal Sample Selection (CromSS) method, a weakly supervised pretraining strategy designed to improve feature representations through cross-modal consistency and noise mitigation techniques. Unlike conventional pretraining approaches, CromSS exploits massive amounts of noisy and easy-to-come-by labels for improved feature learning beneficial to semantic segmentation tasks. We investigate middle and late fusion strategies to optimize the multi-modal pretraining architecture design. We also introduce a crossmodal sample selection module to mitigate the adverse effects of label noise, which employs a cross-modal entangling strategy to refine the estimated confidence masks within each modality to guide the sampling process. Additionally, we introduce a spatialtemporal label smoothing technique to counteract overconfidence for enhanced robustness against noisy labels. To validate our approach, we assembled the multi-modal dataset, NoLDO-S12, which consists of a large-scale noisy label subset from Google's Dynamic World (DW) dataset for pretraining and two downstream subsets with high-quality labels from Google DW and OpenStreetMap (OSM) for transfer learning. Experimental results on two downstream tasks and the publicly available DFC2020 dataset demonstrate that when effectively utilized, the low-cost noisy labels can significantly enhance feature learning for segmentation tasks. The data, codes, and pretrained weights are freely available at https://github.com/zhu-xlab/CromSS.

For agents

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

get_harvested_code_for_paper("2405.01217")
get_code_for_paper("2405.01217")
have("2405.01217")

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

curl https://syntology.ai/api/ran/2405.01217.json

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

[![Syntology run record](https://syntology.ai/api/ran/2405.01217.svg)](https://syntology.ai/paper/2405.01217)

Connect an agent — have() is free.