SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2601.17657 · 2026

SPACE-CLIP: Spatial Perception via Adaptive CLIP Embeddings for Monocular Depth Estimation

Andrew Choi, Taewan Cho, Taeryang Kim

arXiv · PDF · Open in the Atlas

Code that ran

4 of 6 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 4 have no contract verdict; 0 failures of our instrument; 2 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/2601.17657.json · badge: /api/ran/2601.17657.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
taewan2002/space-clip community
Papers with Code archive
4 ran · 2 no recorded run
FunctionStatusWhere it lives
MainDecoderBlock Ran
ran on a synthesized input
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("6c27fd1aef97d491")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 4, 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 torch
import torch.nn as nn
import torch.nn.functional as F

class ResidualConvBlock(nn.Module):

    def __init__(self, num_features, dropout_rate=0.1):
        super().__init__()
        self.conv1 = nn.Conv2d(num_features, num_features, 3, 1, 1, bias=False)
        self.norm1 = nn.BatchNorm2d(num_features)
        self.relu = nn.ReLU(inplace=True)
        self.dropout = nn.Dropout(dropout_rate)
        self.conv2 = nn.Conv2d(num_features, num_features, 3, 1, 1, bias=False)
        self.norm2 = nn.BatchNorm2d(num_features)

    def forward(self, x):
        residual = x
        x = self.relu(self.norm1(self.conv1(x)))
        x = self.dropout(x)
        x = self.norm2(self.conv2(x))
        x += residual
        return self.relu(x)

class UpsampleBlock(nn.Module):

    def __init__(self, in_channels, out_channels):
        super().__init__()
        self.upsample = nn.ConvTranspose2d(in_channels, out_channels, kernel_size=2, stride=2)
        self.conv = nn.Sequential(nn.Conv2d(out_channels, out_channels, 3, padding=1, bias=False), nn.BatchNorm2d(out_channels), nn.ReLU(inplace=True))

    def forward(self, x):
        return self.conv(self.upsample(x))

class MainDecoderBlock(nn.Module):

    def __init__(self, low_res_ch, skip_clip_ch, skip_early_refined_ch, out_ch, dropout=0.1, **kwargs):
        super().__init__()
        self.upsample_low = UpsampleBlock(low_res_ch, out_ch)
        self.skip_clip_gate = nn.Sequential(nn.Conv2d(out_ch, skip_clip_ch, kernel_size=1, bias=True), nn.Sigmoid())
        self.skip_struct_gate = None
        if skip_early_refined_ch > 0:
            self.skip_struct_gate = nn.Sequential(nn.Conv2d(out_ch, skip_early_refined_ch, kernel_size=1, bias=True), nn.Sigmoid())
        self.fusion_conv = nn.Sequential(nn.Conv2d(out_ch + skip_clip_ch + skip_early_refined_ch, out_ch, 3, padding=1, bias=False), nn.BatchNorm2d(out_ch), nn.ReLU(inplace=True), ResidualConvBlock(out_ch, dropout))

    def forward(self, low_res_feat, skip_feat_clip, skip_feat_early_refined):
        low_res_up = self.upsample_low(low_res_feat)
        target_size = low_res_up.shape[-2:]
        if skip_feat_clip.shape[-2:] != target_size:
            skip_feat_clip = F.interpolate(skip_feat_clip, size=target_size, mode='bilinear', align_corners=False)
        if skip_feat_early_refined.shape[-2:] != target_size:
            skip_feat_early_refined = F.interpolate(skip_feat_early_refined, size=target_size, mode='bilinear', align_corners=False)
        clip_gate = self.skip_clip_gate(low_res_up)
        skip_feat_clip = skip_feat_clip * (1.0 + clip_gate)
        if self.skip_struct_gate is not None and skip_feat_early_refined.shape[1] > 0:
            struct_gate = self.skip_struct_gate(low_res_up)
            skip_feat_early_refined = skip_feat_early_refined * (1.0 + struct_gate)
        combined_feat = torch.cat([low_res_up, skip_feat_clip, skip_feat_early_refined], dim=1)
        return self.fusion_conv(combined_feat)


INIT_ARGS = {
    'low_res_ch': 8,
    'skip_clip_ch': 6,
    'skip_early_refined_ch': 4,
    'out_ch': 12,
    'dropout': 0.1,
}

ARGS = {
    'low_res_feat': torch.randn(2, 8, 4, 5),
    'skip_feat_clip': torch.randn(2, 6, 8, 10),
    'skip_feat_early_refined': torch.randn(2, 4, 8, 10),
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
ResidualConvBlock Ran
ran on a synthesized input · constructed an object; no result was computed
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("c94b76202340adbb")
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 an instance of its own class, ResidualConvBlock: the run constructed the object and did not call it, so no result was computed. 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)

# ResidualConvBlock expects:
# - num_features: int (number of input/output channels)
# - dropout_rate: float (optional, default 0.1)

ARGS = {
    "num_features": 16,
    "dropout_rate": 0.1,
}
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 ResidualConvBlock — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a ResidualConvBlock 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
StructuralPathwayBlock Ran
ran on a synthesized input · constructed an object; no result was computed
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("f456f47c622775d3")
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 an instance of its own class, StructuralPathwayBlock: the run constructed the object and did not call it, so no result was computed. 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 sample input tensors
# low_res_feat: (batch=2, channels=8, height=4, width=4)
low_res_feat = torch.randn(2, 8, 4, 4)

# skip_feat_early: (batch=2, channels=16, height=8, width=8)
# Different spatial dimensions to test interpolation
skip_feat_early = torch.randn(2, 16, 8, 8)

ARGS = {
    'low_res_ch': 8,
    'skip_early_ch': 16,
    'out_ch': 12,
}
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 StructuralPathwayBlock — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a StructuralPathwayBlock 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
UpsampleBlock Ran
ran on a synthesized input · constructed an object; no result was computed
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("8df912190eb94c8d")
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 an instance of its own class, UpsampleBlock: the run constructed the object and did not call it, so no result was computed. 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)

# UpsampleBlock takes in_channels and out_channels as parameters
# Looking at the forward method, it expects an input tensor x
# The upsample is ConvTranspose2d which expects 4D input (batch, channels, height, width)
# The conv operations also work on 4D tensors

# Create a sample input tensor: batch_size=2, in_channels=4, height=8, width=8
x = torch.randn(2, 4, 8, 8)

ARGS = {
    'in_channels': 4,
    'out_channels': 3,
}
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 UpsampleBlock — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a UpsampleBlock 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
FiLMLayer Not yet run
unverified (harvested, no recorded run)
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("56646f2666dbc91b")
How it ran

no invocation recorded.

SPACECLIP Not yet run
unverified (harvested, no recorded run)
taewan2002/space-clip/space_clip.py
code served (permissive licence) · get_code("5af31288dfb4c014")
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, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 6 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.

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

Robotic and autonomous systems need dense spatial cues, but many monocular depth models are heavy, task-specific, or hard to attach to an existing multimodal stack. CLIP offers strong semantic representations, yet most CLIP-based depth methods still depend on text prompts or backbone updates, which complicate deployment in integrated control pipelines. We present SPACE-CLIP, a decoder-only depth framework that reads geometric cues directly from a frozen CLIP vision encoder and bypasses the text encoder at inference time. The model combines FiLM-conditioned semantic features from deep layers with structural features from shallow layers to recover both global scene layout and local geometric detail. Under the TFI-FB constraint (text-free inference and frozen vision backbone), SPACE-CLIP achieves AbsRel 0.0901 on KITTI and 0.1042 on NYU Depth V2, and the same dual-pathway decoder transfers to a frozen SigLIP backbone with comparable results. These findings show that a compact decoder can turn a shared foundationmodel backbone into a reusable spatial perception module for embodied AI and autonomous robotic systems. Our model is available at https://github.com/taewan2002/space-clip

For agents

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

get_harvested_code_for_paper("2601.17657")
get_code_for_paper("2601.17657")
have("2601.17657")

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

curl https://syntology.ai/api/ran/2601.17657.json

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

[![Syntology run record](https://syntology.ai/api/ran/2601.17657.svg)](https://syntology.ai/paper/2601.17657)

Connect an agent — have() is free.