SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2103.14304 · 2021

Exploiting Temporal Contexts with Strided Transformer for 3D Human Pose Estimation

arXiv · PDF · Open in the Atlas

Code that ran

8 of 10 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 3 have no contract verdict; 5 failures 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/2103.14304.json · badge: /api/ran/2103.14304.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
Vegetebird/StridedTransformer-Pose3D canonical (classifier label)
classifier basis: census
8 ran · 2 no recorded run
FunctionStatusWhere it lives
camera_to_world Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Vegetebird/StridedTransformer-Pose3D/common/camera.py
code served (permissive licence) · get_code("0892bc8406224611")
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 (2, 5, 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
import torch

torch.manual_seed(0)

# Analyze the function signature and body:
# camera_to_world(X, R, t)
# - X: used in X.shape[:-1] and passed to qrot, so needs shape (..., 3)
# - R: passed to qrot via np.tile(R, (*X.shape[:-1], 1)), needs shape (..., 4) for quaternion
# - t: added to result of qrot, needs shape compatible with qrot output

# qrot(q, v) expects:
# - q: shape (..., 4) for quaternion
# - v: shape (..., 3) for 3D vector
# - output: shape (..., 3)

# X should be a batch of 3D points
X = np.random.randn(2, 5, 3).astype(np.float32)  # batch shape (2, 5), 3D vectors

# R should be a quaternion (4,) that will be tiled to match X's batch dimensions
R = np.array([1.0, 0.0, 0.0, 0.0], dtype=np.float32)  # identity quaternion

# t should be a translation vector compatible with output shape (2, 5, 3)
t = np.random.randn(2, 5, 3).astype(np.float32)

ARGS = {
    "X": X,
    "R": R,
    "t": t
}
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 ndarray, ndim 3, dtype family float, shape (2, 5, 3) — shape (2, 5, 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
coco_h36m Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Vegetebird/StridedTransformer-Pose3D/demo/lib/preprocess.py
code served (permissive licence) · get_code("f3a0f7cf8690505f")
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 numpy as np

# Create a small valid keypoints array for COCO format
# COCO has 17 keypoints, temporal dimension is arbitrary
# keypoints shape: (temporal, num_keypoints, 2) where 2 is (x, y)
temporal = 3  # small temporal dimension
num_keypoints = 17  # COCO standard
keypoints = np.random.rand(temporal, num_keypoints, 2).astype(np.float32) * 100

ARGS = {
    "keypoints": keypoints,
}
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 — 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
deterministic_random Ran
ran on a synthesized input
Vegetebird/StridedTransformer-Pose3D/common/utils.py
code served (permissive licence) · get_code("ec58bab810b80366")
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 hashlib

def deterministic_random(min_value, max_value, data):
    digest = hashlib.sha256(data.encode()).digest()
    raw_value = int.from_bytes(digest[:4], byteorder='little', signed=False)
    return int(raw_value / (2 ** 32 - 1) * (max_value - min_value)) + min_value

ARGS = {
    "min_value": 1,
    "max_value": 100,
    "data": "test_seed_string"
}
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
h36m_coco_format Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Vegetebird/StridedTransformer-Pose3D/demo/lib/preprocess.py
code served (permissive licence) · get_code("85be84dc21eb6ce3")
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 numpy as np

# Create sample keypoints and scores for testing h36m_coco_format
# keypoints shape: (batch_size, temporal, num_joints, 2)
# scores shape: (batch_size, temporal, num_joints)

batch_size = 2
temporal = 5
num_joints = 17  # COCO has 17 joints

# Create keypoints with shape (2, 5, 17, 2)
keypoints = np.random.rand(batch_size, temporal, num_joints, 2).astype(np.float32) * 100

# Create scores with shape (2, 5, 17)
scores = np.random.rand(batch_size, temporal, num_joints).astype(np.float32)

# Ensure some non-zero data
keypoints[0, 0, 1:5, :] = np.array([[10, 20], [15, 25], [12, 22], [18, 28]], dtype=np.float32)
keypoints[0, 0, 5:7, :] = np.array([[30, 35], [32, 37]], dtype=np.float32)
keypoints[0, 0, 11:13, :] = np.array([[50, 55], [52, 57]], dtype=np.float32)

scores[0, 0, :] = np.array([0.9, 0.8, 0.85, 0.88, 0.87, 0.9, 0.89, 0.91, 0.85, 0.86, 0.87, 0.88, 0.89, 0.9, 0.91, 0.92, 0.93], dtype=np.float32)

ARGS = {
    "keypoints": keypoints,
    "scores": scores
}
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 — 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
mpjpe_cal Ran
ran on a synthesized input
Vegetebird/StridedTransformer-Pose3D/common/utils.py
code served (permissive licence) · get_code("c5148aea3c7b76b6")
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 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 torch

torch.manual_seed(0)

# Create two tensors with the same shape
# Shape: (batch_size=2, num_joints=3, coordinates=4)
# This represents 2 samples, 3 joints each, with 4 coordinate values per joint
predicted = torch.randn(2, 3, 4)
target = torch.randn(2, 3, 4)

ARGS = {
    "predicted": predicted,
    "target": target,
}
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 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
revise_kpts Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Vegetebird/StridedTransformer-Pose3D/demo/lib/preprocess.py
code served (permissive licence) · get_code("b9f70678ddc9add2")
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 4, dtype family float, shape (2, 5, 17, 2) 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

# Construct arguments for revise_kpts function
# The function expects:
# - h36m_kpts: array of shape (batch, frames, joints, coords)
# - h36m_scores: array of shape (batch, frames, joints) with confidence scores
# - valid_frames: list of frame indices for each batch item

# Create small test data
batch_size = 2
total_frames = 5
num_joints = 17
coord_dims = 2

# h36m_kpts: keypoints with shape (batch, frames, joints, coords)
h36m_kpts = np.random.randn(batch_size, total_frames, num_joints, coord_dims).astype(np.float32)

# h36m_scores: confidence scores with shape (batch, frames, joints)
# Mix some scores below 0.3 threshold to trigger the conditional logic
h36m_scores = np.random.uniform(0.4, 1.0, (batch_size, total_frames, num_joints)).astype(np.float32)
# Add some low scores to trigger frame corrections
h36m_scores[0, 1, 2] = 0.2  # low score for joint 2
h36m_scores[0, 1, 3] = 0.2  # low score for joint 3
h36m_scores[0, 1, 5] = 0.2  # low score for joint 5
h36m_scores[0, 1, 6] = 0.2  # low score for joint 6
h36m_scores[1, 2, 3] = 0.1  # low score for joint 3
h36m_scores[1, 2, 6] = 0.1  # low score for joint 6

# valid_frames: list of arrays indicating which frames are valid for each batch item
valid_frames = [
    np.array([0, 1, 2, 3, 4]),  # all frames valid for batch 0
    np.array([0, 1, 2, 3, 4])   # all frames valid for batch 1
]

ARGS = {
    "h36m_kpts": h36m_kpts,
    "h36m_scores": h36m_scores,
    "valid_frames": valid_frames
}
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 4, dtype family float, shape (2, 5, 17, 2) — shape (2, 5, 17, 2) 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
show2Dpose Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
Vegetebird/StridedTransformer-Pose3D/demo/vis.py
code served (permissive licence) · get_code("912a1369d1563d3b")
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 int, shape (480, 640, 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 cv2
import numpy as np

# Create a small test image (480x640x3 BGR format)
img = np.ones((480, 640, 3), dtype=np.uint8) * 255

# Create keypoints for 17 joints (standard pose format)
# Each keypoint is (x, y) coordinate
kps = np.array([
    [100, 150],    # 0: nose
    [110, 140],    # 1: left eye
    [120, 135],    # 2: left ear
    [130, 138],    # 3: left shoulder
    [90, 140],     # 4: right eye
    [80, 135],     # 5: right ear
    [70, 138],     # 6: right shoulder
    [100, 200],    # 7: left hip
    [100, 250],    # 8: spine/center
    [100, 300],    # 9: left knee
    [100, 350],    # 10: left ankle
    [100, 300],    # 11: right knee
    [100, 350],    # 12: right ankle
    [100, 300],    # 13: right foot
    [100, 300],    # 14: right hip
    [100, 350],    # 15: right foot
    [100, 350],    # 16: right foot (extra joint)
], dtype=np.float32)

ARGS = {
    "kps": kps,
    "img": img
}
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 int, shape (480, 640, 3) — shape (480, 640, 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
test_calculation Ran
ran on a synthesized input
Vegetebird/StridedTransformer-Pose3D/common/utils.py
code served (permissive licence) · get_code("d8222b9b5b8b97c9")
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 numpy as np

torch.manual_seed(0)

class AverageMeter:
    def __init__(self):
        self.val = 0
        self.avg = 0
        self.sum = 0
        self.count = 0
    
    def update(self, val, n=1):
        self.val = val
        self.sum += val * n
        self.count += n
        self.avg = self.sum / self.count

# Create test tensors
batch_size = 2
num_joints = 4
joint_dims = 3

predicted = torch.randn(batch_size, num_joints, joint_dims)
target = torch.randn(batch_size, num_joints, joint_dims)

# Create action list
action = ["walking 1", "walking 2"]

# Create error_sum dict with nested structure
error_sum = {
    "walking": {
        "p1": AverageMeter(),
        "p2": AverageMeter()
    },
    "running": {
        "p1": AverageMeter(),
        "p2": AverageMeter()
    }
}

data_type = "test"
subject = "S1"

ARGS = {
    "predicted": predicted,
    "target": target,
    "action": action,
    "error_sum": error_sum,
    "data_type": data_type,
    "subject": subject
}
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 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
normalize_screen_coordinates Not yet run
unverified (harvested, no recorded run)
Vegetebird/StridedTransformer-Pose3D/common/camera.py
pointer only (licence: MIT); the stored copy get_code serves, from JihuaPeng/KTPFormer, is pointer only (licence: NONE) · get_code("82baf6aa4fb040a1")
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 JihuaPeng/KTPFormer, and GitHub detected no licence in that repository when it was checked. 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. No invocation recorded.

world_to_camera Not yet run
unverified (harvested, no recorded run)
Vegetebird/StridedTransformer-Pose3D/common/camera.py
code served (permissive licence) · get_code("32eca53ede0ea499")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 9 are also present in other papers whose code we hold (4 with at least one paper dated earlier, 0 the same month, 9 later) and 1 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
normalize_screen_coordinates 10 other papers
2 dated earlier, 8 later
1904.03345 (2019-04)
Semantic Graph Convolutional Networks for 3D Human Pose Regression
2004.13985 (2020-04)
Motion Guided 3D Pose Estimation from Videos
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
and 7 more not named here
camera_to_world 4 other papers
4 later
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
2206.06420 (2022-06)
GraphMLP: A Graph MLP-Like Architecture for 3D Human Pose Estimation
Li_Hourglass_Tokenizer_for_Efficient_Transformer-Based_3D_Human_Pose_Estimation_CVPR_2024_paper (2024 (year only))
Hourglass Tokenizer for Efficient Transformer-Based 3D Human Pose Estimation
and 1 more not named here
world_to_camera 4 other papers
1 dated earlier, 3 later
2004.13985 (2020-04)
Motion Guided 3D Pose Estimation from Videos
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
2206.06420 (2022-06)
GraphMLP: A Graph MLP-Like Architecture for 3D Human Pose Estimation
and 1 more not named here
deterministic_random 3 other papers
1 dated earlier, 2 later
2004.13985 (2020-04)
Motion Guided 3D Pose Estimation from Videos
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
aaai_28549 (2024 (year only))
Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation
mpjpe_cal 3 other papers
1 dated earlier, 2 later
2004.13985 (2020-04)
Motion Guided 3D Pose Estimation from Videos
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
aaai_28549 (2024 (year only))
Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation
coco_h36m 2 other papers
2 later
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
2310.16288 (2023-10)
MotionAGFormer: Enhancing 3D Human Pose Estimation with a Transformer-GCNFormer Network
h36m_coco_format 2 other papers
2 later
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
2310.16288 (2023-10)
MotionAGFormer: Enhancing 3D Human Pose Estimation with a Transformer-GCNFormer Network
revise_kpts 2 other papers
2 later
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
2310.16288 (2023-10)
MotionAGFormer: Enhancing 3D Human Pose Estimation with a Transformer-GCNFormer Network
test_calculation 2 other papers
2 later
2111.12707 (2021-11)
MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation
aaai_28549 (2024 (year only))
Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation

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

Despite the great progress in 3D human pose estimation from videos, it is still an open problem to take full advantage of a redundant 2D pose sequence to learn representative representations for generating one 3D pose. To this end, we propose an improved Transformer-based architecture, called Strided Transformer, which simply and effectively lifts a long sequence of 2D joint locations to a single 3D pose. Specifically, a Vanilla Transformer Encoder (VTE) is adopted to model long-range dependencies of 2D pose sequences. To reduce the redundancy of the sequence, fully-connected layers in the feed-forward network of VTE are replaced with strided convolutions to progressively shrink the sequence length and aggregate information from local contexts. The modified VTE is termed as Strided Transformer Encoder (STE), which is built upon the outputs of VTE. STE not only effectively aggregates long-range information to a single-vector representation in a hierarchical global and local fashion, but also significantly reduces the computation cost. Furthermore, a full-to-single supervision scheme is designed at both full sequence and single target frame scales applied to the outputs of VTE and STE, respectively. This scheme imposes extra temporal smoothness constraints in conjunction with the single target frame supervision and hence helps produce smoother and more accurate 3D poses. The proposed Strided Transformer is evaluated on two challenging benchmark datasets, Human3.6M and HumanEva-I, and achieves state-of-the-art results with fewer parameters. Code and models are available at \url{https://github.com/Vegetebird/StridedTransformer-Pose3D}.

For agents

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

get_harvested_code_for_paper("2103.14304")
get_code_for_paper("2103.14304")
have("2103.14304")

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

curl https://syntology.ai/api/ran/2103.14304.json

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

[![Syntology run record](https://syntology.ai/api/ran/2103.14304.svg)](https://syntology.ai/paper/2103.14304)

Connect an agent — have() is free.