Pith. sign in

REVIEW 5 major objections 6 minor 48 references

V-MIND: Building Versatile Monocular Indoor 3D Detector with Diverse 2D Annotations

T0 review · 5 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read V-MIND shows that large-scale 2D datasets, lifted into 3D with monocular depth and camera calibration, can train a versatile indoor 3D object detector that detects new classes without any 3D labels.

desk verdict V-MIND's lifted-2D-data approach to monocular indoor 3D detection is plausible and worth a serious look; pseudo-box quality and narrow comparisons need work, not rejection. read the letter →

arxiv 2412.11412 v1 pith:X2R4WPLH submitted 2024-12-16 cs.CV

classification cs.CV
keywords monocular3Dobjectdetectionindoorsceneunderstandingpseudo-labelgeneration2D-to-3Dliftingself-calibrationlossambiguitylarge-vocabularyCLIPembeddings
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

The paper claims that the scarcity of 3D annotations, not the detector architecture, is the main bottleneck for versatile monocular indoor 3D detection. To remove it, V-MIND converts large-scale 2D instance masks from LVIS into pseudo 3D bounding boxes by lifting pixels with a monocular metric depth estimator and a camera intrinsic predictor. A self-calibration loss lets the detector learn a per-image scale correction for the noisy lifted boxes, and an ambiguity loss prevents the original 3D dataset from penalizing correct detections of classes it never annotated. Trained jointly on real 3D data and the lifted pseudo data, V-MIND improves AP2D from 9.64 to 12.98 and AP3D from 7.52 to 8.52 on Omni3DIN, and it detects 38 new classes that the baseline cannot see at all. This points to a path for scaling 3D detectors to many object classes without additional 3D annotation.

What carries the argument

The load-bearing pieces are a pixel lifting pipeline, a self-calibration loss, and an ambiguity loss. The lifting pipeline back-projects each 2D instance mask into a 3D point cloud via $[x,y,z]^\top = ZK^{-1}[u,v,1]^\top$ using predicted depth $Z$ and predicted camera intrinsics $K$, then fits a tight oriented 3D bounding box after outlier removal and per-class point-count filtering. The self-calibration loss applies a learnable per-image scale $\exp(s)$ to the predicted box center before computing Chamfer distance against the pseudo box, regularized by $\|s\|_1$, which lets the detector absorb global depth errors during training. The ambiguity loss $-\log \sum_{c \in \Omega_{new} \cup \Omega_{bkg}} \Pr(c)$ treats the union of new classes and background as one group, avoiding incorrect penalties when the 3D dataset lacks annotations for objects that the 2D data says are real. Classification is performed in a pre-trained CLIP embedding space so semantically related class names share feature structure.

What would settle it

On ScanNet200, compute the ratio of predicted to ground-truth depth for each lifted object; if the within-image variance of these ratios is comparable to the between-image variance, a single per-image scale cannot absorb the errors, and the pseudo-box signal for new classes would be unreliable.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that pseudo 3D boxes lifted from 2D-only annotations are a viable training signal for monocular 3D object detection, provided the detector can correct a per-image scale error and is shielded from missing-annotation ambiguity. On the Omni3D indoor benchmark, adding lifted LVIS data raises overall AP2D from 9.64 to 12.98 and AP3D from 7.52 to 8.52, with new classes reaching AP2D 5.89 and AP3D 2.66 despite having no 3D ground truth. The small drop in original-class AP3D (14.39 versus the baseline 15.04) is the price paid for this broader vocabulary, and the paper argues the trade is favorable overall.

Load-bearing premise

The depth errors in the lifted pseudo boxes are assumed to be approximately correctable by a single per-image scale, so that after calibration the boxes are a trustworthy training signal for classes that have no real 3D labels.

Editorial extensions

If this is right

  • Adding more 2D labeled images directly enlarges the pseudo 3D training set, so the detector should keep improving as 2D data scales up.
  • Better monocular depth estimators and camera intrinsic predictors would translate into better pseudo boxes and higher AP3D, especially for new classes.
  • The self-calibration and ambiguity losses are detector-agnostic, so the same recipe could be applied to other monocular 3D detectors such as BEV-based ones.
  • Classifying in a pre-trained vision-language space becomes more important as the vocabulary grows, and should help with long-tailed and semantically related classes.

Reading between the lines

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

  • The single per-image scale $\exp(s)$ is a coarse model of depth error; a natural extension would be per-object or spatially varying scale corrections, which could recover some of the 4.3% drop on original classes.
  • The ambiguity loss is a general remedy for partially labeled detection data and could transfer to 2D open-vocabulary detection where background proposals may contain unannotated novel objects.
  • The low pseudo-box precision on ScanNet200 (29.72%) suggests that filtering by predicted confidence or by agreement between multiple depth models could yield a higher-quality training set with little recall loss.
  • Since new-class detection relies entirely on pseudo boxes, the method's ceiling for novel classes is set by the depth and intrinsic models; combining it with self-supervised depth finetuning on the target domain is a testable way to push APnew3D higher.
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.

Referee Report

5 major / 6 minor

Summary. The paper proposes V-MIND, a monocular indoor 3D object detector that augments the Cube R-CNN architecture with pseudo 3D training data lifted from the large-vocabulary 2D dataset LVIS. Lifting uses an off-the-shelf metric depth estimator (ZoeDepth) and a camera intrinsic predictor (Wild Camera), followed by point-cloud denoising, per-category point thresholds, and oriented 3D box fitting. The detector is trained jointly on Omni3DIN and the lifted boxes using three additional components: a CLIP embedding classifier, a self-calibration loss that applies a per-image scalar scale to predicted box centers when supervising on pseudo boxes, and an ambiguity loss for new classes that are present in images of the 3D dataset but unannotated there. On Omni3DIN, the method reports AP2D 12.98 vs. 9.64 and AP3D 8.52 vs. 7.52 over Cube R-CNN, with nonzero APnew3D 2.66 for new classes.

Significance. If the reported results hold, the paper offers a practical recipe for expanding the class vocabulary of monocular 3D detectors without additional 3D annotation, which is a meaningful step toward scalable indoor 3D perception. The paper has several strengths: the ablations in Table 3 isolate the contributions of the two proposed losses, the pseudo-box quality is explicitly evaluated on ScanNet200 (Table 1), and the per-category filtering thresholds are derived from ScanNet200 and CLIP embeddings rather than from the target Omni3D evaluation set, which reduces circularity. The CLIP-based classification is well motivated for the long-tailed class distribution. However, the strength of the empirical claims is tempered by the low reported pseudo-box precision (29.72%), the training-only use of the self-calibration network, and the absence of comparisons to state-of-the-art monocular 3D detectors other than Cube R-CNN variants.

major comments (5)
  1. [§4.2, Table 1] The pseudo-box evaluation on ScanNet200 reports precision 29.72% and recall 16.53% without stating the matching protocol: it is unclear whether these numbers use 3D IoU, which IoU threshold (if any), and how predicted boxes are matched to ground-truth boxes. Since new classes receive all of their 3D supervision from these pseudo boxes (Section 3.1, Table 3), this omission is load-bearing. Please specify the protocol and report precision/recall at several 3D IoU thresholds (e.g., 0.1, 0.25, 0.5), ideally per class for the 38 new classes, so the reader can judge how many pseudo boxes are usable for 3D localization rather than only for 2D region supervision.
  2. [§3.3, Eq. (3)] The self-calibration mechanism applies a single per-image scalar exp(s) only to the predicted box center x; it cannot correct spatially varying depth errors, errors in box dimensions w,h,l, or orientation R(p). Moreover, the text states that the calibration network is applied only for pseudo 3D data during training, so at inference the detector outputs the uncalibrated x while the training loss optimizes exp(s)x. This creates a potential train/test mismatch, and the concern is directly relevant to the Table 3 ablation where the self-calibration loss raises APnew3D from 1.31 to 2.66. Please clarify whether the calibration network is used at inference, or provide evidence that the training-only use does not harm inference, e.g., by reporting the learned distribution of s, the residual depth error after calibration, or an experiment that applies the calibration network at inference.
  3. [§4.2, Table 2] The paper claims state-of-the-art performance, but the only comparisons are Cube R-CNN and Cube R-CNN with CLIP embeddings. Section 2.2 acknowledges that UniMode outperforms Cube R-CNN on Omni3D, yet no quantitative comparison to UniMode or to any other published monocular 3D detector on Omni3DIN is provided. Please either add comparisons to published Omni3DIN results or soften the claim to 'improves over Cube R-CNN' throughout the abstract and conclusion.
  4. [§4.1, Original class and New class] The selection of the 38 new classes is described only as 'excluding those with relatively few instances in the LVIS dataset.' The paper does not report which classes are excluded, the LVIS instance counts per new class, or per-class AP. The aggregate APnew3D of 2.66 could be driven by a small number of easy classes. Please provide the full list of new classes, their LVIS instance counts, and per-class AP2D/AP3D, or at least a histogram of per-class AP, so the versatility claim is supported.
  5. [§4.2, Table 2] All results appear to be single runs without error bars or multiple seeds. Given that the AP3D improvement is 8.52 vs. 7.52 and APoriginal3D drops by 4.3%, the paper should report variance over at least three seeds or state that the differences are stable across runs; without this, the reader cannot assess whether the headline gains are within noise.
minor comments (6)
  1. [Table 1 caption] The caption contains a typo: 'Quantitaive' should be 'Quantitative'.
  2. [Abstract and §1] The phrase 'state-of-the-art' appears in the abstract and introduction, but the experimental section compares only against Cube R-CNN baselines; this should be reconciled with comment 3 above.
  3. [§4.1] The paper states that Omni3DIN contains 84 classes but only 38 are used as original classes and 38 as new classes, totaling 76; please clarify the relationship between the 84 classes and the 76 classes used in training.
  4. [§4.2, Table 2] The use of '∞' for relative improvement over a zero baseline is unconventional and potentially misleading; please define the notation more explicitly or report absolute improvements.
  5. [§3.4, Eq. (5)] The ambiguity loss sums probabilities over all new classes plus background for every background-assigned proposal; this may over-encourage new-class probability for proposals that are truly background. Please discuss this trade-off or report an ablation with a limited candidate set.
  6. [§4.2, Figure 4] The caption claims the lifted boxes are 'accurate in terms of categories, locations, and orientations,' which is in tension with the 29.72% precision in Table 1; please soften the caption or add quantitative support.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation identified: the Omni3DIN gains are evaluated against manual 3D ground truth, with pseudo-label thresholds from ScanNet200 and CLIP, and all corrective losses are train-time only.

full rationale

I walked the paper's derivation chain: pseudo 3D boxes are generated from LVIS masks, ZoeDepth depth estimates, and Wild Camera intrinsics; per-category point thresholds are computed from ScanNet200 and CLIP embeddings (Section 3.1), not from Omni3D evaluation; the detector is trained jointly on Omni3D 3D annotations and the lifted pseudo boxes; the self-calibration and ambiguity losses (Sections 3.3 and 3.4) are train-time objectives; and evaluation is on Omni3DIN with AP computed over IoU thresholds 0.05-0.50 against manual ground-truth boxes. No fitted constant is taken from the target test set, and none of the proposed loss functions changes the definition of the benchmark metric. The only candidate reduction would be if the new-class predictions were simply echoing LVIS masks, but the model is evaluated on Omni3D ground-truth boxes it never sees during training, and the pseudo boxes are auxiliary training targets rather than the evaluation target. The low ScanNet200 precision (29.72%) reported in Table 1 is a correctness and label-quality concern, not a circularity concern, because it does not make the evaluation quantity equal to an input of the method. The load-bearing external components (Cube R-CNN, Omni3D, ZoeDepth, Wild Camera, CLIP) are cited from independent prior work, with no uniqueness theorem or central premise resting on a self-citation chain. Accordingly, the central claim is self-contained with respect to circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced. The ledger is dominated by dataset-derived thresholds and assumptions about external pretrained models, which are standard for a training-pipeline paper.

free parameters (3)
  • Per-category point thresholds for pseudo-box filtering = Statistics derived from ScanNet200 object point counts, not numerically reported
    Section 3.1 filters out pseudo 3D boxes with too few points using a class-specific threshold computed from ScanNet200; the exact threshold values are not listed.
  • Outlier removal radius and neighbor count = Unspecified
    Section 3.1 removes points with few neighbors within a given radius or distant from nearest neighbors, but the radius and neighbor-count values are not reported.
  • Self-calibration regularization weight = Unspecified
    The L1 regularization term in Eq. (4) controls how much the model can rely on the per-image scale exp(s), but its loss weight is not stated.
assumptions (4)
  • domain assumption Monocular metric depth and camera intrinsic predictors produce sufficiently accurate estimates for pseudo-label generation.
    Section 3.1 relies on ZoeDepth [2] and the Wild Camera predictor [47]; if their errors are large or biased, the lifted pseudo boxes are unreliable.
  • domain assumption LVIS instance masks are accurate enough to define clean object point clouds.
    Section 3.1 uses LVIS masks directly for lifting; mask errors propagate into the pseudo 3D boxes.
  • domain assumption CLIP text embedding cosine similarity is a valid proxy for semantic similarity between object categories.
    Section 3.1 maps new categories to ScanNet200 classes using CLIP embeddings to derive point thresholds, and Section 3.2 uses CLIP embeddings as the classifier space.
  • domain assumption Omni3D ground truth covers only a fixed set of classes, and any object from a new class is labeled as background unless addressed.
    Section 3.4 introduces the ambiguity loss to handle proposals on new-class objects that the original 3D dataset treats as background.

how reviews work

0 comments
Cite this review

Pith. "Pith review of V-MIND: Building Versatile Monocular Indoor 3D Detector with Diverse 2D Annotations." pith.science (2026). https://pith.science/paper/X2R4WPLH

@misc{pith2026241211412,
  author       = {Pith},
  title        = {Pith review of: V-MIND: Building Versatile Monocular Indoor 3D Detector with Diverse 2D Annotations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X2R4WPLH}},
  note         = {Machine review of arXiv:2412.11412}
}
read the original abstract

The field of indoor monocular 3D object detection is gaining significant attention, fueled by the increasing demand in VR/AR and robotic applications. However, its advancement is impeded by the limited availability and diversity of 3D training data, owing to the labor-intensive nature of 3D data collection and annotation processes. In this paper, we present V-MIND (Versatile Monocular INdoor Detector), which enhances the performance of indoor 3D detectors across a diverse set of object classes by harnessing publicly available large-scale 2D datasets. By leveraging well-established monocular depth estimation techniques and camera intrinsic predictors, we can generate 3D training data by converting large-scale 2D images into 3D point clouds and subsequently deriving pseudo 3D bounding boxes. To mitigate distance errors inherent in the converted point clouds, we introduce a novel 3D self-calibration loss for refining the pseudo 3D bounding boxes during training. Additionally, we propose a novel ambiguity loss to address the ambiguity that arises when introducing new classes from 2D datasets. Finally, through joint training with existing 3D datasets and pseudo 3D bounding boxes derived from 2D datasets, V-MIND achieves state-of-the-art object detection performance across a wide range of classes on the Omni3D indoor dataset.

Figures

Figures reproduced from arXiv: 2412.11412 by the authors.

Figure 1
Figure 1. Augmented 2D-to-3D dataset generation. Given a 2D image dataset with diverse object classes, we first use a metric-based depth estimator to predict the depth map and an intrinsic predictor to derive the camera’s intrinsic parameters. We then compute and denoise the object point cloud for each instance and generate a 3D bounding box (Section 3.1) that tightly encloses the point cloud. 2.3. Large and Open Vocabulary O… view at source ↗
Figure 2
Figure 2. Overview of the proposed V-MIND. (a) Given an input image, a backbone model extracts features, which are then used to generate region proposals. Each proposal is classified in a pre-trained vision-language space, with a 3D bounding box predicted by the cube head (Section 3.2). Leveraging an augmented 2D-to-3D dataset, our model can detect not only the original object classes but also previously unseen ones, i.e., th… view at source ↗
Figure 3
Figure 3. Example to illustrate the ambiguity issue. Data from the 3D dataset with a limited vocabulary treats proposals enclosing new-class objects as background (e.g., proposal B in the figure), which interferes with the training of the 3D detector and hinders its accuracy in identifying new classes. To this end, we propose the ambiguity loss Lamb. Since a proposal assigned as background in the 3D dataset (Omni3D) could bel… view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results of the 3D pseudo bounding boxes lifted from the LVIS [11] 2D dataset. The proposed 3D data generation method can produce accurate 3D bounding boxes in terms of categories, locations, and orientations. Monocular 3D Detection. We show our detection ac…
Figure 5
Figure 5. Figure 5: Qualitative results of V-MIND and baseline Cube R-CNN on (a) SUN-RGBD, (b) ARKitScenes, and (c) Hypersim. As demonstrated in (a), V-MIND effectively predicts new–class objects—specifically, a monitor and keyboard in the first row and a bag in the second row–while Cube …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages

  1. [1]

    ARK- itscenes - a diverse real-world dataset for 3d indoor scene understanding using mobile RGB-d data

    Gilad Baruch, Zhuoyuan Chen, Afshin Dehghan, Tal Dimry, Yuri Feigin, Peter Fu, Thomas Gebauer, Brandon Joffe, Daniel Kurz, Arik Schwartz, and Elad Shulman. ARK- itscenes - a diverse real-world dataset for 3d indoor scene understanding using mobile RGB-d data. In Thirty-fifth NeurIPS Datasets and Benchmarks Track (Round 1) , 2021. 2, 6

  2. [2]

    Zoedepth: Zero-shot trans- fer by combining relative and metric depth

    Shariq Farooq Bhat, Reiner Birkl, Diana Wofk, Peter Wonka, and Matthias M ¨uller. Zoedepth: Zero-shot trans- fer by combining relative and metric depth. arXiv preprint arXiv:2302.12288, 2023. 1, 3, 7

  3. [3]

    Omni3d: A large benchmark and model for 3d object detection in the wild

    Garrick Brazil, Abhinav Kumar, Julian Straub, Nikhila Ravi, Justin Johnson, and Georgia Gkioxari. Omni3d: A large benchmark and model for 3d object detection in the wild. In IEEE Conference on Computer Vision and Pattern Recog- nition (CVPR), 2023. 1, 2, 4, 5, 6, 7

  4. [4]

    Coda: Col- laborative novel box discovery and cross-modal alignment for open-vocabulary 3d object detection

    Yang Cao, Yihan Zeng, Hang Xu, and Dan Xu. Coda: Col- laborative novel box discovery and cross-modal alignment for open-vocabulary 3d object detection. In Advances in Neural Information Processing Systems (NeurIPS), 2023. 1, 3

  5. [5]

    End-to- end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers. In European Confer- ence on Computer Vision (ECCV), 2020. 2

  6. [6]

    Exploring classification equilibrium in long-tailed object detection

    Chengjian Feng, Yujie Zhong, and Weilin Huang. Exploring classification equilibrium in long-tailed object detection. In IEEE International Conference on Computer Vision (ICCV),

  7. [7]

    Dqs3d: Densely-matched quantization- aware semi-supervised 3d detection

    Huan-ang Gao, Beiwen Tian, Pengfei Li, Hao Zhao, and Guyue Zhou. Dqs3d: Densely-matched quantization- aware semi-supervised 3d detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 21905–21915, 2023. 3

  8. [8]

    Fast r-cnn

    Ross Girshick. Fast r-cnn. In IEEE International Conference on Computer Vision (ICCV), 2015. 2

Show all 48 references
  1. [9]

    Rich feature hierarchies for accurate object detection and semantic segmentation

    Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. Rich feature hierarchies for accurate object detection and semantic segmentation. In IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2014. 2

  2. [10]

    Open-vocabulary object detection via vision and language knowledge distillation

    Xiuye Gu, Tsung-Yi Lin, Weicheng Kuo, and Yin Cui. Open-vocabulary object detection via vision and language knowledge distillation. In International Conference on Learning Representations (ICLR), 2022. 3

  3. [11]

    LVIS: A dataset for large vocabulary instance segmentation

    Agrim Gupta, Piotr Dollar, and Ross Girshick. LVIS: A dataset for large vocabulary instance segmentation. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019. 1, 2, 3, 5, 6

  4. [12]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Gir- shick. Mask r-cnn. In IEEE International Conference on Computer Vision (ICCV), 2017. 2

  5. [13]

    Cooperative holistic scene understanding: Unifying 3d object, layout and camera pose estimation

    Siyuan Huang, Siyuan Qi, Yinxue Xiao, Yixin Zhu, Ying Nian Wu, and Song-Chun Zhu. Cooperative holistic scene understanding: Unifying 3d object, layout and camera pose estimation. In Advances in Neural Information Process- ing Systems (NeurIPS), 2018. 2

  6. [14]

    3d-relnet: Joint object and relational network for 3d prediction

    Nilesh Kulkarni, Ishan Misra, Shubham Tulsiani, and Abhi- nav Gupta. 3d-relnet: Joint object and relational network for 3d prediction. In IEEE International Conference on Com- puter Vision (ICCV), pages 2212 – 2221, October 2019. 2

  7. [15]

    Cornernet: Detecting objects as paired keypoints

    Hei Law and Jia Deng. Cornernet: Detecting objects as paired keypoints. In European Conference on Computer Vi- sion (ECCV), 2018. 2

  8. [16]

    Overcoming classifier im- balance for long-tail object detection with balanced group softmax

    Yu Li, Tao Wang, Bingyi Kang, Sheng Tang, Chunfeng Wang, Jintao Li, and Jiashi Feng. Overcoming classifier im- balance for long-tail object detection with balanced group softmax. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2020. 3

  9. [17]

    Unimode: Unified monocular 3d object detection

    Zhuoling Li, Xiaogang Xu, SerNam Lim, and Hengshuang Zhao. Unimode: Unified monocular 3d object detection. arXiv preprint arXiv:2402.18573, 2024. 2

  10. [18]

    Ssd: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. Ssd: Single shot multibox detector. In European Con- ference on Computer Vision (ECCV), 2016. 2

  11. [19]

    Open-vocabulary point-cloud object detection without 3d an- notation

    Yuheng Lu, Chenfeng Xu, Xiaobao Wei, Xiaodong Xie, Masayoshi Tomizuka, Kurt Keutzer, and Shanghang Zhang. Open-vocabulary point-cloud object detection without 3d an- notation. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2023. 1

  12. [20]

    Open-vocabulary point-cloud object detection without 3d an- notation

    Yuheng Lu, Chenfeng Xu, Xiaobao Wei, Xiaodong Xie, Masayoshi Tomizuka, Kurt Keutzer, and Shanghang Zhang. Open-vocabulary point-cloud object detection without 3d an- notation. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2023. 3

  13. [21]

    Total3dunderstanding: Joint lay- out, object pose and mesh reconstruction for indoor scenes from a single image

    Yinyu Nie, Xiaoguang Han, Shihui Guo, Yujian Zheng, Jian Chang, and Jian Jun Zhang. Total3dunderstanding: Joint lay- out, object pose and mesh reconstruction for indoor scenes from a single image. In IEEE Conference on Computer Vi- sion and Pattern Recognition (CVPR), 2020. 2

  14. [22]

    On model calibration for long-tailed object detection and instance segmentation

    Tai-Yu Pan, Cheng Zhang, Yandong Li, Hexiang Hu, Dong Xuan, Soravit Changpinyo, Boqing Gong, and Wei-Lun Chao. On model calibration for long-tailed object detection and instance segmentation. In Advances in Neural Informa- tion Processing Systems (NeurIPS), 2021. 3

  15. [23]

    High quality entity segmentation

    Lu Qi, Jason Kuen, Tiancheng Shen, Jiuxiang Gu, Wenbo Li, Weidong Guo, Jiaya Jia, Zhe Lin, and Ming-Hsuan Yang. High quality entity segmentation. In IEEE International Conference on Computer Vision (ICCV), 2023. 4

  16. [24]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. arXiv preprint arXiv:2103.00020, 2021. 1, 3, 4, 6

  17. [25]

    Improved visual-semantic alignment for zero-shot object detection

    Shafin Rahman, Salman Khan, and Nick Barnes. Improved visual-semantic alignment for zero-shot object detection. In AAAI Conference on Artificial Intelligence (AAAI), 2020. 3

  18. [26]

    You only look once: unified, real-time object detec- tion (2015)

    Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: unified, real-time object detec- tion (2015). arXiv preprint arXiv:1506.02640, 2015. 2

  19. [27]

    Yolo9000: Better, faster, stronger

    Joseph Redmon and Ali Farhadi. Yolo9000: Better, faster, stronger. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2017. 3 9

  20. [28]

    Faster r-cnn: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. Advances in Neural Information Pro- cessing Systems (NeurIPS), 2015. 2

  21. [29]

    Susskind

    Mike Roberts, Jason Ramapuram, Anurag Ranjan, Atulit Kumar, Miguel Angel Bautista, Nathan Paczan, Russ Webb, and Joshua M. Susskind. Hypersim: A photorealistic syn- thetic dataset for holistic indoor scene understanding. In IEEE International Conference on Computer Vision (ICCV),

  22. [30]

    Language- grounded indoor 3d semantic segmentation in the wild

    David Rozenberszki, Or Litany, and Angela Dai. Language- grounded indoor 3d semantic segmentation in the wild. In European Conference on Computer Vision (ECCV), 2022. 4, 6

  23. [31]

    Imvoxelnet: Image to voxels projection for monocular and multi-view general-purpose 3d object detection

    Danila Rukhovich, Anna V orontsova, and Anton Konushin. Imvoxelnet: Image to voxels projection for monocular and multi-view general-purpose 3d object detection. In Winter Conference on Applications of Computer Vision (WACV) ,

  24. [32]

    Imvoxelnet: Image to voxels projection for monocular and multi-view general-purpose 3d object detection

    Danila Rukhovich, Anna V orontsova, and Anton Konushin. Imvoxelnet: Image to voxels projection for monocular and multi-view general-purpose 3d object detection. In Winter Conference on Applications of Computer Vision (WACV) , pages 2397–2406, January 2022. 2

  25. [33]

    Object detection with trans- formers: A review, 2023

    Tahira Shehzadi, Khurram Azeem Hashmi, Didier Stricker, and Muhammad Zeshan Afzal. Object detection with trans- formers: A review, 2023. 2

  26. [34]

    Bharat Singh, Hengduo Li, Abhishek Sharma, and Larry S. Davis. R-fcn-3000 at 30fps: Decoupling detection and clas- sification. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2018. 3

  27. [35]

    Lichtenberg, and Jianxiong Xiao

    Shuran Song, Samuel P. Lichtenberg, and Jianxiong Xiao. Sun rgb-d: A rgb-d scene understanding benchmark suite. In IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2015. 2, 6

  28. [36]

    Equalization loss v2: A new gradient balance ap- proach for long-tailed object detection

    Jingru Tan, Xin Lu, Gang Zhang, Changqing Yin, and Quan- quan Li. Equalization loss v2: A new gradient balance ap- proach for long-tailed object detection. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 3

  29. [37]

    Equalization loss for long-tailed object recognition

    Jingru Tan, Changbao Wang, Buyu Li, Quanquan Li, Wanli Ouyang, Changqing Yin, and Junjie Yan. Equalization loss for long-tailed object recognition. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020. 3

  30. [38]

    Im- geonet: Image-induced geometry-aware voxel representation for multi-view 3d object detection

    Tao Tu, Shun-Po Chuang, Yu-Lun Liu, Cheng Sun, Ke Zhang, Donna Roy, Cheng-Hao Kuo, and Min Sun. Im- geonet: Image-induced geometry-aware voxel representation for multi-view 3d object detection. In IEEE International Conference on Computer Vision (ICCV), 2023. 2

  31. [39]

    Efros, and Jitendra Malik

    Shubham Tulsiani, Saurabh Gupta, David Fouhey, Alexei A. Efros, and Jitendra Malik. Factoring shape, pose, and layout from the 2d image of a 3d scene. In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018. 2

  32. [40]

    Seesaw loss for long- tailed instance segmentation

    Jiaqi Wang, Wenwei Zhang, Yuhang Zang, Yuhang Cao, Jiangmiao Pang, Tao Gong, Kai Chen, Ziwei Liu, Chen Change Loy, and Dahua Lin. Seesaw loss for long- tailed instance segmentation. In IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2021. 3

  33. [41]

    Detecting 11k classes: Large scale object detection without fine-grained bounding boxes

    Hao Yang, Hao Wu, and Hao Chen. Detecting 11k classes: Large scale object detection without fine-grained bounding boxes. In IEEE International Conference on Computer Vi- sion (ICCV), 2019. 3

  34. [42]

    Open-vocabulary object detection using captions

    Alireza Zareian, Kevin Dela Rosa, Derek Hao Hu, and Shih- Fu Chang. Open-vocabulary object detection using captions. In IEEE Conference on Computer Vision and Pattern Recog- nition (CVPR), 2021. 3

  35. [43]

    Distribution alignment: A unified framework for long-tail visual recognition

    Songyang Zhang, Zeming Li, Shipeng Yan, Xuming He, and Jian Sun. Distribution alignment: A unified framework for long-tail visual recognition. In IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2021. 3

  36. [44]

    Detecting twenty-thousand classes using image-level supervision

    Xingyi Zhou, Rohit Girdhar, Armand Joulin, Philipp Kr¨ahenb¨uhl, and Ishan Misra. Detecting twenty-thousand classes using image-level supervision. In European Confer- ence on Computer Vision (ECCV), 2022. 3

  37. [45]

    Ob- jects as points

    Xingyi Zhou, Dequan Wang, and Philipp Kr ¨ahenb¨uhl. Ob- jects as points. arXiv preprint arXiv:1904.07850, 2019. 2

  38. [46]

    On the continuity of rotation representations in neural networks

    Yi Zhou, Connelly Barnes, Jingwan Lu, Jimei Yang, and Hao Li. On the continuity of rotation representations in neural networks. In IEEE Conference on Computer Vision and Pat- tern Recognition (CVPR), 2019. 4

  39. [47]

    Tame a wild camera: in-the-wild monocular camera calibra- tion

    Shengjie Zhu, Abhinav Kumar, Masa Hu, and Xiaoming Liu. Tame a wild camera: in-the-wild monocular camera calibra- tion. Advances in Neural Information Processing Systems , 36, 2024. 1, 3, 7

  40. [48]

    Deformable {detr}: Deformable transform- ers for end-to-end object detection

    Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable {detr}: Deformable transform- ers for end-to-end object detection. In International Confer- ence on Learning Representations (ICLR), 2021. 2 10

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.