Refine3DNet: Scaling Precision in 3D Object Reconstruction from Multi-View RGB Images using Attention

Ajith Balakrishnan; Linu Shine; Sreeja S

REVIEW 5 major objections 5 minor 32 references

Refine3DNet: Scaling Precision in 3D Object Reconstruction from Multi-View RGB Images using Attention

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Refine3DNet claims a 4.2% IoU gain in single-view 3D reconstruction on ShapeNet.

desk verdict The Refiner module as specified cannot upsample from 4^3 to 32^3, so the reported SOTA IoU numbers are untestable; the rest of the paper never recovers. read the letter →

arxiv 2412.00731 v1 pith:ASI6DM7F submitted 2024-12-01 cs.CV

Ajith Balakrishnan , Sreeja S , Linu Shine This is my paper · ORCID

classification cs.CV

keywords 3Dreconstructionmulti-viewself-attentionvoxelShapeNetautoencoderrefinernetworkJTSO

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Refine3DNet claims to reconstruct high-quality 3D voxel models from single or multiple 2D RGB images by combining a ResNet-based autoencoder with a transformer-inspired self-attention module and a 3D U-Net refiner. The paper reports state-of-the-art mean intersection-over-union scores on the ShapeNet test set, outperforming the previous best by 4.2% in single-view reconstruction. The authors also introduce a three-phase training scheme, Joint Train Separate Optimization (JTSO), which separates encoder-decoder learning from attention learning. The motivation is that existing methods struggle to fuse features from unordered multi-view images and degrade when only one image is available.

What carries the argument

The architecture is an encoder-decoder with a self-attention bottleneck and a separate 3D U-Net refiner. The shared encoder is a 12-layer ResNet that turns each 127×127 RGB image into a 1024-d feature vector; the multi-head self-attention (eight heads, scaled dot-product) fuses those vectors for arbitrary numbers of views; the decoder produces an initial 32³ voxel grid; and the 3D U-Net refiner, with U-Net skip connections, refines that grid. Training uses mean per-voxel cross-entropy loss, and the JTSO algorithm updates the encoder-decoder, attention module, and refiner in three phases to avoid degradation when input count changes.

What would settle it

Compute the output spatial size of the refiner's decoder: with input 4³, kernel 4, padding 2, and stride 1, each transposed convolution yields output size 4×1 - 2×2 + 4 = 4, so the final output remains 4³, not the claimed 32³. Running the described refiner on any 32³ input would reveal the dimension mismatch.

Watch

Extended reading notes

Core claim

The paper's central claim is that its hybrid architecture, trained with JTSO, achieves the highest mean IoU on ShapeNet for both single-view and multi-view voxel reconstruction, with a single-view mean IoU of 0.689 versus 0.661 for the next-best Pix2Vox-A. For up to 15 views the proposed network holds the lead, and at 20 views it trails EVolT by only 0.002. The authors attribute the gains to self-attention's ability to aggregate features from unordered image sets and to the 3D U-Net refiner that cleans the initial decoder output.

Load-bearing premise

The paper assumes the 3D U-Net refiner can take a 32³ voxel grid, compress it through three 2×2×2 max-poolings to 4³, and then expand it back to 32³ using transposed convolutions with stride 1, which would leave the spatial size at 4³ rather than restoring 32³.

Editorial extensions

If this is right

If the reported IoU gains hold, single-image 3D reconstruction becomes accurate enough for practical use in VR, robotics, and medical imaging.
The JTSO training schedule could let other multi-view networks train more robustly with variable numbers of input views.
Self-attention over unordered latent feature sets is a viable alternative to RNN-based fusion, avoiding permutation variance.
The 4.2% improvement is most pronounced in categories like lamp and chair, suggesting attention helps most on objects with fine or complex geometry.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The reported 20-view performance trailing EVolT hints that the attention module's advantage shrinks as views increase; a testable extension is to blend the two fusion strategies.
The refiner's dimension inconsistency suggests that the actual implementation likely includes an implicit upsampling or a different stride; if the refiner were removed, the decoder's raw output might already carry most of the reported gain.
JTSO's three-phase scheme could be applied to other encoder-decoder architectures with auxiliary 3D heads, decoupling representation learning from fusion learning.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Let me cut to the chase: the central claim of Refine3DNet is not supportable as written. The Refiner module described in Section 3.4 takes a 32^3 voxel volume, runs three stride-2 conv+pool blocks down to 4^3, then decodes with three transposed convolutions explicitly set to stride 1 'to maintain spatial resolution.' That gives a 4^3 output, yet the loss and IoU compare against 32^3 ground truth. The paper even declares 'input and output dimensions of the Refiner are 32^3.' That is a load-bearing internal contradiction, and no code or pretrained model is provided to resolve it. So the headline number (mean IoU 0.689, 4.2% over Pix2Vox-A) does not follow from the architecture described.

There are other signs of carelessness. Equation (5) for IoU has the same expression in numerator and denominator. Table 1 lists VolT and EVolT with identical per-category values in 11 of 13 rows, which looks like a copy-paste error rather than a real result. The JTSO training schedule is described vaguely—three phases with separate parameter updates—but no ablation isolates its contribution, and the refiner's impact is only shown in a single figure, not in the final tables. The resource table claims 143M parameters and 2900MB memory, which is higher than all baselines, but the paper does not discuss whether the gains justify the cost. The introduction also has a malformed citation '(author?)[5]', which signals sloppy revision.

To be fair, the paper is not nonsense. The architecture is a plausible combination of known pieces: ResNet18 encoder, transformer-style self-attention over the 1024-d latent, decoder to 32^3, and a 3D U-Net refiner. That is a reasonable recipe, and the authors clearly know the related literature, including AttSets, Pix2Vox, VolT, and EVolT. The multi-view table shows the proposed network trailing EVolT at 20 views (0.733 vs 0.735), which is oddly honest, and the claim of strength at low views is consistent with what a refiner would do. The JTSO idea—decoupling encoder-decoder, attention, and refiner training—is worth exploring, though it is underspecified here.

But a paper whose central result depends on a refiner that cannot exist as specified, with a wrong metric equation and suspicious baseline tables, does not deserve referee time. The reader who wants to try the recipe should wait for code or a corrected version. My recommendation: desk reject, with an invitation to resubmit a corrected and code-backed version.

Referee Report

5 major / 5 minor

Summary. The paper proposes Refine3DNet, a hybrid CNN-transformer architecture for single- and multi-view 3D voxel reconstruction from RGB images. The pipeline consists of a ResNet-based encoder, a self-attention feature aggregation module, a decoder that produces an initial 32^3 occupancy volume, and a 3D U-Net-style refiner, trained with a proposed three-stage Joint Train Separate Optimization (JTSO) procedure. The authors report state-of-the-art mean IoU on a ShapeNet benchmark, including 0.689 for single-view reconstruction, surpassing the next best method by 4.2%. The claimed contribution is an improved architecture plus a training strategy that decouples encoder-decoder, attention, and refiner updates.

Significance. If the reported numbers were credible, this would be a useful empirical result: a relatively simple hybrid architecture outperforming strong baselines on a standard public benchmark, with an interesting training decoupling idea (JTSO) and a comparison across 1-20 input views. The paper explicitly compares against 3D-R2N2, AttSets, Pix2Vox, VolT, and EVolT on the public ShapeNet test setup, which is the right protocol for benchmarking. However, the central construction contains multiple internal inconsistencies: the refiner cannot change spatial resolution as described, the IoU formula in Eq. (5) is mathematically wrong, the loss in Eq. (4) has the wrong sign, and Table 1 contains suspicious duplicated baseline entries. These issues directly affect every reported accuracy number, and no code, pretrained models, or detailed ablations are provided to resolve them. The claimed state of the art is therefore not established.

major comments (5)

[Section 3.4] The Refiner module as specified cannot map a 32^3 input to a 32^3 output. Three 2x2x2 max-pooling layers after 4x4x4 convolutions with padding 2 reduce the spatial size from 32 to 4, while the decoder uses transposed convolutions with 'a stride of 1 to maintain spatial resolution', which would leave the volume at 4^3. The paper states that input and output dimensions of the Refiner are 32^3 and uses 32^3 ground truth in the loss and IoU, so the reported refinement gains are not realizable from the given architecture. This is a load-bearing internal inconsistency, not a stylistic issue, and no code or pretrained model is supplied to disambiguate it.
[Section 4.2, Eq. (5)] The IoU formula is incorrect: the denominator is identical to the numerator, so the metric is always 1 for any nonempty prediction if the threshold is exceeded anywhere in the ground truth, and undefined otherwise. The correct denominator should be the size of the union of the predicted and ground-truth voxel sets. Since all quantitative comparisons in Tables 1-2 and Figure 6 are reported as IoU, the metric error invalidates the reported numbers as stated.
[Section 4.1, Eq. (4)] The loss function lacks a minus sign. As written, L is the sum of non-positive terms, so lower values mean worse agreement with the ground truth, contrary to the text's statement that 'a lower value of L signifies a prediction that is more accurately aligned with the ground truth.' The training objective is therefore described with the opposite optimization direction, which affects the validity of the reported training procedure.
[Table 1] In Table 1, VolT and EVolT have identical per-category IoU values in 11 of 13 categories, differing only in Rifle and Sofa, yet their overall means are reported as 0.605 and 0.609, respectively. This pattern is not explained and suggests copying or transcription errors in the baseline numbers. Because the headline 4.2% improvement is computed relative to these baselines, the comparison table needs to be corrected and independently verified.
[Section 6.1 and Section 5] The claimed benefit of the refiner is supported only by Figure 7, which shows a plot without numerical values, confidence intervals, or a table of the underlying numbers. Given the architectural problem in Section 3.4, the refiner's contribution cannot be assessed from the manuscript. An ablation with exact IoU values for the decoder-only output and the refiner output is necessary to support the central claim.

minor comments (5)

[Section 3.3] The text says the decoder consists of five residual blocks but then lists six output filter sizes (128, 128, 128, 64, 64, 32); please correct the count or the list.
[Section 6.2] In the discussion of Table 3, 'For inference, the backward pass time is 78.12 ms' should presumably read 'forward pass time'; backward pass is not an inference operation.
[Table 1] The row label 'Table' is confusing because it is simultaneously a category name and the name of the table; consider using 'Table (furniture)' or a clearer caption.
[References] Reference [30] is incomplete: it lists only authors 'K. Hwang and W. Sung' with no title, venue, or year.
[Equation (1)] The notation for the scaling factor in the softmax attention is typeset inconsistently as 'sqrt(d_k)' with an unusual radical symbol; please use the standard form for readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an empirical benchmark against the external ShapeNet test set, with no fitted parameter or self-citation chain doing load-bearing work.

full rationale

The paper's central claim is comparative: Refine3DNet with JTSO reports mean IoU 0.689 on single-view and 0.733 at 20-view ShapeNet reconstruction, versus external baselines (3D-R2N2, Attsets, Pix2Vox, VolT, EVolT). These are empirical evaluations against a public dataset, not derived quantities that reduce to the paper's own inputs. The loss (Eq. 4) and IoU metric (Eq. 5) compare predicted and ground-truth voxel occupancy, and the baselines are independently published methods. I found no self-definitional construction, no fitted parameter renamed as a prediction, no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation. The paper does contain a serious architectural inconsistency: the Refiner encoder's three 2x2x2 max-pooling operations reduce a 32^3 input to 4^3, while the decoder's stride-1 transposed convolutions cannot restore 32^3, so the reported refiner-based numbers do not follow from the architecture as written. That is an internal correctness and reproducibility problem, not circularity, because the claimed SOTA comparison still rests on external data and baselines rather than on a definitional equivalence. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The paper does not introduce new physical entities; its contributions are a network architecture and a training schedule. The core results rest on standard deep learning components, public dataset assumptions, and the questionable architectural premise that the refiner can upsample as described.

free parameters (3)

IoU binarization threshold = 0.25
Used to convert predicted occupancy probabilities into binary voxels before computing IoU. The reported scores are sensitive to this threshold, and no sensitivity analysis is provided.
Number of self-attention heads = 8
Multi-head attention parameter chosen by hand; standard in transformers but not justified for this task and not ablated.
Learning rate schedule = 0.001, decay factor 2 after 150 epochs
Optimizer hyperparameters chosen without a search; the central result depends on them.

assumptions (3)

domain assumption The ShapeNet subset, renderings, and 80:20 split used here match the baselines' evaluation setups.
Baseline IoU values are quoted from prior papers; the paper does not show that the test inputs are identical across methods.
ad hoc to paper The Refiner decoder's stride-1 transposed convolutions can recover the original 32^3 volume after three 2x2x2 max-pooling operations.
Section 3.4 states the decoder uses stride 1 to maintain spatial resolution, which cannot undo the encoder's 8x downsampling. The architecture description is internally inconsistent.
domain assumption The printed cross-entropy loss in Eq. (4) and IoU in Eq. (5) are typos, and the standard versions were actually used.
As written, Eq. (4) is a positive log-loss and Eq. (5) yields IoU = 1; the paper's numbers make sense only if the standard definitions were intended.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Refine3DNet: Scaling Precision in 3D Object Reconstruction from Multi-View RGB Images using Attention." pith.science (2026). https://pith.science/paper/ASI6DM7F

@misc{pith2026241200731,
  author       = {Pith},
  title        = {Pith review of: Refine3DNet: Scaling Precision in 3D Object Reconstruction from Multi-View RGB Images using Attention},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ASI6DM7F}},
  note         = {Machine review of arXiv:2412.00731}
}

read the original abstract

Generating 3D models from multi-view 2D RGB images has gained significant attention, extending the capabilities of technologies like Virtual Reality, Robotic Vision, and human-machine interaction. In this paper, we introduce a hybrid strategy combining CNNs and transformers, featuring a visual auto-encoder with self-attention mechanisms and a 3D refiner network, trained using a novel Joint Train Separate Optimization (JTSO) algorithm. Encoded features from unordered inputs are transformed into an enhanced feature map by the self-attention layer, decoded into an initial 3D volume, and further refined. Our network generates 3D voxels from single or multiple 2D images from arbitrary viewpoints. Performance evaluations using the ShapeNet datasets show that our approach, combined with JTSO, outperforms state-of-the-art techniques in single and multi-view 3D reconstruction, achieving the highest mean intersection over union (IOU) scores, surpassing other models by 4.2% in single-view reconstruction.

Figures

Figures reproduced from arXiv: 2412.00731 by the authors.

**Figure 2.** Detailed architecture of the proposed network [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Illustration of multi-head attention and Scaled dot [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 5.** Figure 5: Proposed JTSO Algorithm flow The proposed network generates two 3D models for each input: one from the encoder-decoder and another from the refiner network. Both models are updated using mean cross-entropy loss compared to ground truth. The training algorithm has thre…

**Figure 6.** Figure 6: Average Intersection over Union (IOU) by Category [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: The evaluation of IoU is influenced by both the [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 26 canonical work pages

[5]

Haozhe Xie, Hongxun Yao, Xiaoshuai Sun, Shangchen Zhou, Shengping Zhang, Xiaojun Tong .Pix2Vox: Context-aware 3D Reconstruction from Single and Multi-view Images CVPR-2019-arXiv:1901.11153

work page Pith review arXiv 2019
[1]

Efficient 3D Object Reconstruction using Visual Transformers

Agarwal, Rohan and Zhou, Wei and Wu, Xiaofeng and Li, Yuhan. "Ef- ficient 3D Object Reconstruction using Visual Transformers" . CVPR (2023). https://doi.org/10.48550/arXiv.2302.08474

work page Pith review arXiv doi:10.48550/arxiv.2302.08474 2023
[2]

Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction

Bo Yang , Sen Wang , Andrew Markham , Niki Trigoni."Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction" . Int J Comput Vis (2019). https://doi.org/10.1007/s11263-019-01217-w

work page doi:10.1007/s11263-019-01217-w 2019
[3]

3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction

Christopher B. Choy, Danfei Xu, JunYoung Gwak Kevin Chen, Silvio Savarese. "3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction" . ECCV 2016,DOI-10.1007/978-3-319-46484-8_38

work page doi:10.1007/978-3-319-46484-8_38 2016
[4]

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative- Adversarial Modeling

jiajun Wu, Chengkai Zhang, Tianfan Xue, William T Freeman, Joshua B Tenen- baum ."Learning a Probabilistic Latent Space of Object Shapes via 3D Generative- Adversarial Modeling",29th NIPS 2016, Barcelona

work page 2016
[6]

U-Net: Convolutional Net- works for Biomedical Image Segmentation

Olaf Ronneberger, Philipp Fischer, and Thomas Brox ."U-Net: Convolutional Net- works for Biomedical Image Segmentation" ., IEEE 2015

work page 2015
[7]

Large- Scale 3D Shape Reconstruction and Segmentation from ShapeNet Core55

Li Yi , Lin Shao , Manolis Savva , Haibin Huang , Yang Zhou , Qirui Wang ,Benjamin Graham , Martin Engelcke , Roman Klokov , Victor Lempitsky , Yuan Gan ." Large- Scale 3D Shape Reconstruction and Segmentation from ShapeNet Core55" . ECCV 2017

work page 2017
[8]

Beyond PASCAL:A benchmark for 3D object detection in the wild. In W ACV 2014

Y. Xiang, R. Mottaghi, and S. Savarese." Beyond PASCAL:A benchmark for 3D object detection in the wild. In W ACV 2014"

work page 2014

Show all 32 references

[9]

A Point Set Generation Network for 3D Object Reconstruction from a Single Image

Haoqiang Fan, Hao Su, and Leonidas Guibas." A Point Set Generation Network for 3D Object Reconstruction from a Single Image ".,2017 IEEE Conference On Computer Vision And Pattern Recognition

2017
[10]

Dense 3D Point Cloud Reconstruction Using a Deep Pyramid Network

Priyanka MandikalR. Venkatesh Babu. " Dense 3D Point Cloud Reconstruction Using a Deep Pyramid Network " .,2019 IEEE Winter Conference On Applications Of Computer Vision

2019
[11]

Learning a Multi-View Stereo Machine

Abhishek Kar, Christian Häne, Jitendra Malik. " Learning a Multi-View Stereo Machine"., NIPS 2017

2017
[12]

MarrNet : 3 D Shape Reconstruction via 2 . 5 D Sketches

Jiajun Wu, Yifan Wang." MarrNet : 3 D Shape Reconstruction via 2 . 5 D Sketches" ., NIPS 2017

2017
[13]

DeepMVS: Learning Multi-view Stereopsis

Po-Han Huang, Kevin Matzen, Johannes Kopf, Narendra Ahuja, Jia-Bin Huang."DeepMVS: Learning Multi-view Stereopsis" ., CVPR 2018

2018
[14]

Past, Present, and Future of Simultaneous Localization and Mapping: To- wards the Robust-Perception Age

C. Cadena, L. Carlone, H. Carrillo, Y. Latif, D. Scaramuzza, J. Neira, I. D. Reid, and J. J. Leonard."Past, Present, and Future of Simultaneous Localization and Mapping: To- wards the Robust-Perception Age. ", IEEE Transactions on Robotics, 32(6):1309–1332, 2016

2016
[15]

A Survey of Structure from Motion

O. Ozyesil, V. Voroninski, R. Basri, and A. Singer. "A Survey of Structure from Motion. ", Acta Numerica, 26:305–364, 2017

2017
[16]

3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annota- tion

Ozgun Cicek, Ahmed Abdulkadir, Soeren S. Lienkamp, Thomas Brox, Olaf Ron- neberger."3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annota- tion. ", MICCAI 2016

2016
[17]

Deep residual learning for image recognition

He, K., Zhang, Ren. S., Sun.j"Deep residual learning for image recognition. ", IEEE conference on computer vision and pattern recognition (pp. 770-778)
[18]

Very deep convolutional networks for large-scale image recognition

K. Simonyan and A. Zisserman.“Very deep convolutional networks for large-scale image recognition. ”[Online]. Available: https://arxiv.org/abs/1409.1556 (Sep, 2014)

2014 arXiv
[19]

International Conference on Neural Information Processing Systems

Zaheer M, Kottur S, Ravanbakhsh S, Poczos B, Salakhutdinov R, Smola A (2017)."Deep Sets. International Conference on Neural Information Processing Systems

2017
[20]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, Attention Is All You Need, arXiv preprint arXiv:1706.03762, 2017

2017 arXiv
[21]

C. B. Choy, D. Xu, J. Gwak, K. Chen, and S. Savarese,3D-R2N2: A unified approach for single and multi-view 3D object reconstruction , in European Conference on Computer Vision 1 (ECCV), pp. 645-661, 2016

2016
[22]

X. Chen, H. Chen, Y. Peng, L. Liu, and C. Huang, A Freehand 3D Ultrasound Reconstruction Method Based on Deep Learning , Electronics, vol. 12, no. 7, p. 1527, 2023, doi: 10.3390/electronics12071527

2023 doi
[23]

Sizikova, X

E. Sizikova, X. Cao, A. Lewis, K. Moise, and M. Coffee,Improving Computed Tomog- raphy (CT) Reconstruction via 3D Shape Induction , arXiv preprint arXiv:2208.10937, 2022

2022 arXiv
[24]

W. Li, H. Yang, Z. Hu, J. Zheng, G.-S. Xia, and C. He,3D Building Reconstruction from Monocular Remote Sensing Images with Multi-level Supervisions, arXiv preprint arXiv:2404.04823, 2024

2024 arXiv
[25]

Zioga, A

M. Zioga, A. Nikopoulou, M. Alexandridi, D. Maintas, M. Mikeli, A.-N. Rapso- manikis, and E. Stiliaris, Image Reconstruction in the Positron Emission Tomography, HNPS Proceedings, vol. 20, pp. 73, 2012, doi: 10.12681/hnps.2490

2012 doi
[27]

Vinyals, S

O. Vinyals, S. Bengio, and M. Kudlur, Order Matters: Sequence to sequence for sets , arXiv preprint arXiv:1511.06391, 2016

2016 arXiv
[28]

Lowe, Distinctive Image Features from Scale-Invariant Keypoints , Inter- national Journal of Computer Vision, vol

D. Lowe, Distinctive Image Features from Scale-Invariant Keypoints , Inter- national Journal of Computer Vision, vol. 60, no. 2, pp. 91-, 2004, doi: 10.1023/B:VISI.0000029664.99615.94

2004
[29]

Çiçek, A

Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox, and O. Ronneberger,3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation , arXiv preprint arXiv:1606.06650, 2016

2016 arXiv
[30]

Hwang and W

K. Hwang and W. Sung,
[31]

Pascanu, T

R. Pascanu, T. Mikolov, and Y. Bengio, On the difficulty of training Recurrent Neural Networks, arXiv preprint arXiv:1211.5063, 2013

2013 arXiv
[32]

D. Wang, X. Cui, X. Chen, Z. Zou, T. Shi, S. Salcudean, Z. J. Wang, and R. Ward, Multi-view 3D Reconstruction with Transformers, in CVPR, 2023 [Online]. Available: http://arxiv.org/abs/2103.12957

2023 arXiv
[33]

Wiles and A

O. Wiles and A. Zisserman, SilNet: Single- and Multi-View Reconstruction by Learning from Silhouettes, arXiv preprint arXiv:1711.07888, 2017

2017 arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [5]

Haozhe Xie, Hongxun Yao, Xiaoshuai Sun, Shangchen Zhou, Shengping Zhang, Xiaojun Tong .Pix2Vox: Context-aware 3D Reconstruction from Single and Multi-view Images CVPR-2019-arXiv:1901.11153

work page Pith review arXiv 2019

[2] [1]

Efficient 3D Object Reconstruction using Visual Transformers

Agarwal, Rohan and Zhou, Wei and Wu, Xiaofeng and Li, Yuhan. "Ef- ficient 3D Object Reconstruction using Visual Transformers" . CVPR (2023). https://doi.org/10.48550/arXiv.2302.08474

work page Pith review arXiv doi:10.48550/arxiv.2302.08474 2023

[3] [2]

Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction

Bo Yang , Sen Wang , Andrew Markham , Niki Trigoni."Attentional Aggregation of Deep Feature Sets for Multi-view 3D Reconstruction" . Int J Comput Vis (2019). https://doi.org/10.1007/s11263-019-01217-w

work page doi:10.1007/s11263-019-01217-w 2019

[4] [3]

3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction

Christopher B. Choy, Danfei Xu, JunYoung Gwak Kevin Chen, Silvio Savarese. "3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction" . ECCV 2016,DOI-10.1007/978-3-319-46484-8_38

work page doi:10.1007/978-3-319-46484-8_38 2016

[5] [4]

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative- Adversarial Modeling

jiajun Wu, Chengkai Zhang, Tianfan Xue, William T Freeman, Joshua B Tenen- baum ."Learning a Probabilistic Latent Space of Object Shapes via 3D Generative- Adversarial Modeling",29th NIPS 2016, Barcelona

work page 2016

[6] [6]

U-Net: Convolutional Net- works for Biomedical Image Segmentation

Olaf Ronneberger, Philipp Fischer, and Thomas Brox ."U-Net: Convolutional Net- works for Biomedical Image Segmentation" ., IEEE 2015

work page 2015

[7] [7]

Large- Scale 3D Shape Reconstruction and Segmentation from ShapeNet Core55

Li Yi , Lin Shao , Manolis Savva , Haibin Huang , Yang Zhou , Qirui Wang ,Benjamin Graham , Martin Engelcke , Roman Klokov , Victor Lempitsky , Yuan Gan ." Large- Scale 3D Shape Reconstruction and Segmentation from ShapeNet Core55" . ECCV 2017

work page 2017

[8] [8]

Beyond PASCAL:A benchmark for 3D object detection in the wild. In W ACV 2014

Y. Xiang, R. Mottaghi, and S. Savarese." Beyond PASCAL:A benchmark for 3D object detection in the wild. In W ACV 2014"

work page 2014