Pith. sign in

REVIEW 4 major objections 5 minor 26 references

Co-Learning: Towards Semi-Supervised Object Detection with Road-side Cameras

T0 review · 4 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 With 10% labeled frames, a teacher–student detector matches fully supervised roadside detection.

desk verdict The paper's central parity claim is undercut by its own inconsistent numbers and a mislabeled oracle baseline. read the letter →

arxiv 2411.19143 v1 pith:3EEP6F36 submitted 2024-11-28 cs.CV

Jicheng Yuan , Anh Le-Tuan , Ali Ganbarov , Manfred Hauswirth , Danh Le-Phuoc This is my paper · ORCID
classification cs.CV
keywords semi-supervisedlearningobjectdetectionteacher-studentnetworkpseudo-labelrefinementannotationalignmentroadsidecamerastext-derivedannotations
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 is trying to establish that semi-supervised learning can make roadside object detection practical when labels are scarce: a teacher–student framework called Co-Learning trains on 10% labeled frames while treating the other 90% as unlabeled, and the authors report that it reaches performance comparable to a fully supervised solution. On the AI City Challenge Track 2 dataset, with Faster R-CNN and a ResNet-50 backbone, the aligned Co-Learning model scores 36.5 mean average precision (mAP), versus 36.1 for an oracle model trained only on pseudo-labels and 23.0 for the same framework without annotation alignment. The paper attributes the gain to attacking label inconsistency: pseudo-label assignment, feature alignment, and threshold refinement stabilize what the student learns, while stemming, lemmatization, and synonym clustering standardize the text-derived box labels. A reader should care because labeling every frame from dense roadside camera networks is expensive, and a method that works with 10% of the labels lowers the barrier to deploying detectors at the edge.

What carries the argument

The load-bearing object is a teacher-student network with mutual learning, in which a teacher generates pseudo-labels from unlabeled frames and a multi-head student trains on them under exponential-moving-average coupling. Three sub-modules carry the method: dynamic pseudo-label assignment, which replaces static IoU matching so noisy box boundaries do not misassign anchors; pseudo-feature alignment, which lets each classification feature pick a suitable regression feature vector; and pseudo-label refinement, which adaptively adjusts the score threshold to keep the pseudo-ground-truth count stable early in training. The supporting machinery is annotation alignment, built from stemming, lemmatization, and semantic-role labeling: free-text vehicle descriptions are reduced to color, type, and motion attributes and then collapsed into a representative class (for example, 'red van', 'van', and 'blue van' all become 'van'). Together these pieces are meant to make the student learn from a consistent, less noisy label set rather than from contradictory pseudo-boxes.

What would settle it

Run the with-alignment and without-alignment models against a single human-verified label set and score both on the same label vocabulary; if the aligned model's advantage over the oracle disappears or shrinks to noise, the reported 36.5 mAP result is an artifact of annotation remapping rather than evidence for the semi-supervised learning method.

Watch

Extended reading notes

Core claim

Co-Learning's central claim is that the bottleneck in semi-supervised roadside detection is label inconsistency rather than labeled-data scarcity. Using 10% of the AI City Challenge Track 2 frames as labeled data and 90% as unlabeled data, a Faster R-CNN teacher-student detector with ResNet-50 achieves 36.5 mAP (AP@0.5) when annotation alignment is applied, exceeding the 36.1 mAP of the oracle model that can access all pseudo-labels; without alignment the same pipeline drops to 23.0 mAP. The authors conclude that consistent pseudo-labels—maintained by dynamic thresholds, cross-teacher/student agreement, and a standardized label vocabulary derived from natural-language descriptions—are what allow a small labeled set to approach supervised-level detection.

Load-bearing premise

The load-bearing premise is that box labels created from natural-language text—via stemming, lemmatization, and synonym clustering—are reliable enough to serve as ground truth for both training and evaluation; if those labels are systematically wrong, or if the two variants are scored on different label vocabularies, the reported improvement is an artifact of relabeling rather than evidence for the semi-supervised method.

Editorial extensions

If this is right

  • If Co-Learning works as reported, roadside deployments could annotate only 10% of frames and still obtain detectors whose mAP is on par with models trained on the complete pseudo-label set.
  • The framework implies that label consistency, not the sheer volume of unlabeled data, is the main lever in semi-supervised roadside object detection.
  • The dynamic assignment and feature-alignment modules should reduce sensitivity to pseudo-box boundary noise, making the method applicable to dense traffic scenes where static IoU assignment is fragile.
  • The authors position the pipeline as ready to transfer to edge devices such as Jetson Orin and Xavier, where the 10%-label regime is practical under limited compute.

Reading between the lines

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

  • The comparison in the paper is against an oracle trained only on pseudo-labels, so the abstract's 'fully supervised' phrasing is stronger than the experiment: a human-verified label ceiling would let the 36.5 mAP result be measured against true full supervision.
  • Part of the 0.4 mAP gain over the oracle may come from the evaluation label vocabulary changing between the aligned and unaligned conditions; scoring both variants on one fixed label set would isolate the learning contribution from the relabeling.
  • A natural extension is to sweep the labeled fraction (for example, 1%, 5%, 20%) to see whether annotation alignment matters more or less as supervision becomes scarcer.
  • The text-standardization recipe is not tied to road scenes and could be probed on other weakly labeled visual domains, such as surveillance or aerial imagery, where natural-language descriptions accompany the frames.
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

4 major / 5 minor

Summary. The paper proposes Co-Learning, a teacher-student semi-supervised object detection framework for roadside camera data, evaluated on the 7th AI City Challenge Track 2 dataset. The method combines three modules—dynamic pseudo-label assignment, pseudo-feature alignment, and pseudo-label refinement—with an annotation-alignment strategy that maps natural-language vehicle descriptions to a smaller set of box-level class labels via stemming, lemmatization, and synonym clustering. Using 10% labeled data, the authors report a mean average precision (mAP) of 36.5 with annotation alignment versus 23.0 without it, and claim that this is comparable to fully-supervised solutions. The central evidence for this claim is Table 1, which compares an 'Oracle' model trained on pseudo-labels with variants with and without annotation alignment.

Significance. If the reported results were reliable, the paper would offer a useful contribution to semi-supervised object detection on edge-roadside cameras, particularly the idea of aligning noisy text-derived annotations into a consistent label taxonomy. However, the evaluation is not currently trustworthy: the headline mAP values in Table 1 are internally inconsistent with the per-class APs, the 'Oracle' baseline is not a fully-supervised detector, the annotation-alignment comparison appears circular with respect to the label space, and no ablations are provided for the three proposed modules. The paper's central claim of parity with fully-supervised solutions is therefore unsupported. The paper does address a relevant practical problem and proposes a coherent pipeline, but the experimental validation as presented is insufficient to establish the claimed contribution.

major comments (4)
  1. [Table 1, Section 3.3] The printed mAP values in Table 1 are inconsistent with the per-class AP values in the same table. Averaging the seven per-class APs gives Oracle 26.1, w/o Annotation-Alignment 16.7, and w/ Annotation-Alignment 27.2, whereas the table reports 36.1, 23.0, and 36.5. Because the caption defines mAP as mean average precision, the headline numbers appear to be computational errors or to rely on an unstated class set or weighting. The 36.5 mAP that anchors the abstract's parity claim therefore cannot be taken at face value.
  2. [Abstract, Section 3.3, Section 3.4] The abstract claims Co-Learning 'achieves comparable performance as fully-supervised solutions using 10% labeled data,' but no fully-supervised baseline is reported. The 'Oracle' row in Table 1 is defined as a model trained only on pseudo-labels, not as a detector trained on human box annotations. Moreover, Section 3.3 states that the aligned model is 'further trained on the fully annotated dataset,' while Section 3.4 states that only 10% of annotations are used; the two statements are contradictory. The parity claim is therefore unsupported and needs either a proper fully-supervised comparison or a revision of the claim.
  3. [Section 3.1, Table 1] The comparison 'w/ Annotation-Alignment' versus 'w/o Annotation-Alignment' is confounded by the label space used for evaluation. The paper applies stemming, lemmatization, and synonym clustering to define a new label taxonomy (e.g., 'red van', 'van', and 'blue van' are all mapped to 'van'), but it does not specify whether both rows are evaluated on the original label space, the aligned label space, or each on its own training label space. If the w/ model is evaluated on the aligned taxonomy while the w/o model is evaluated on the original taxonomy, the reported improvement is an artifact of label remapping. A fixed held-out label space for evaluation must be used.
  4. [Section 3.3] The paper attributes the result to three modules (dynamic pseudo-label assignment, pseudo-feature alignment, and pseudo-label refinement) but provides no ablation or individual evaluation of these modules. Without ablations, the contribution of each module to the reported mAP cannot be assessed, and the central method claim is not substantiated.
minor comments (5)
  1. [Table 1 caption, Section 3.4] The caption states 'We reported AP0.5 as the mean averaged precision mAP,' but Section 3.4 refers simply to mAP; please define the evaluation protocol precisely, including the IoU threshold, the class set, and the averaging method.
  2. [Section 3.1] The subscript formatting in the standardized text format is corrupted in the manuscript; the definition of the three attribute parts should be presented readably.
  3. [Section 2.2] The heading 'Semi-supverised Object Detection' contains a typo.
  4. [Table 1, Section 3.3] No error bars, confidence intervals, or multiple-seed results are reported; the 0.4% improvement over the Oracle is small enough that it could arise from randomness, so an uncertainty estimate is needed.
  5. [Section 3.4] The dataset split is underspecified: it is not stated how the 10% labeled subset was selected, how many images or frames are used, or how the text descriptions are paired with video frames.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 23.0→36.5 mAP gain is substantially definitional: 'annotation alignment' is a synonym-clustering label remap, and the reported mAP is computed on the remapped label vocabulary, so the improvement is at least partly forced by the label-space change rather than by the SSL method.

  1. self definitional [Section 3.1 (Data Curation and Annotation Alignment) and Section 3.3 / Table 1]
    "To ensure consistent box-level annotation and minimize the complexity of the learning aim, we then grouped these synonyms into clusters based on semantic similarity and replaced them with a representative term. For instance, terms such as red van, van, and blue van are grouped under the label van. ... When further trained on the fully annotated dataset with additional unlabeled data and the annotation alignment strategy, the performance escalates to an mAP of 36.5."

    The 'annotation alignment' whose benefit is claimed is exactly the synonym clustering that defines the evaluation vocabulary. Merging distinct labels (e.g., red van, blue van, van) into one class (van) reduces the number of classes and removes fine-grained confusions, so mean AP on the merged vocabulary rises even if detections are unchanged. The paper compares w/ and w/o alignment across this label-space change without fixing the evaluation class set; the improvement is therefore partly a consequence of the remapping, not an independent prediction of the SSL framework. The internal inconsistency of Table 1 (printed mAPs 23.0 and 36.5 are not the means of the listed per-class APs, which are 16.7 and 27.2) reinforces that the mAP values are not a controlled, reproducible comparison.

full rationale

The significant circularity is confined to the annotation-alignment evaluation. The paper's own Section 3.1 defines the aligned label vocabulary; Section 3.3 then credits the alignment with the large mAP jump. Because mAP is computed over that vocabulary, the comparison with the unaligned setting does not isolate the SSL method's contribution. No self-citation chain is load-bearing: references to prior tracked-vehicle-retrieval work ([19], [20]) motivate the stemming/lemmatization but are not the source of the central claim. The teacher-student pseudo-label loop is standard SSOD practice and, by itself, is not circular since evaluation is on held-out labels. Separate concerns are evidentiary rather than circular: the abstract's 'fully-supervised' parity claim is not backed by any fully-supervised baseline (the 'Oracle' is 'trained only on pseudo labels'), and Table 1's printed mAPs do not match the per-class APs. These problems lower confidence in the claim but do not themselves constitute a circular derivation; the partial circularity score of 6 reflects the definitional role of the label remapping in the reported improvement.

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

The central claim rests on a hand-chosen label taxonomy, undisclosed pseudo-label thresholds, and hyperparameters fitted via Bayesian optimization. It also assumes that text-derived annotations are valid ground truth and that standard SSOD assumptions transfer to this dataset without validation. The three named modules are introduced without formal specification, making them non-independent inventions that cannot be verified.

free parameters (3)
  • Annotation alignment label taxonomy = Seven classes: Sedan, Bus, Pickup-Truck, SUV, Hatchback, Van, Truck
    The synonym clusters are hand-defined in Section 3.1 and directly determine the classes used in training and in the mAP computation; the comparison between 'w/' and 'w/o' annotation alignment is therefore sensitive to this choice.
  • Dynamic pseudo-label threshold parameters = Not reported
    The pseudo-label refinement module adjusts score thresholds dynamically (Section 3.3), but no thresholds, bounds, or schedules are given; these settings materially affect mAP.
  • Hyperparameter search results = Not reported
    Section 3.2 states Bayesian Optimization was used over hyperparameters, meaning final hyperparameters were fitted on the validation data; the fitted values are not disclosed.
assumptions (3)
  • domain assumption Text descriptions in AI City Challenge Track 2 can be reliably converted to box-level annotations using SRL, stemming, and lemmatization.
    Section 3.1 builds the training/evaluation labels from text; no validation of these derived annotations against human ground truth is provided.
  • domain assumption Teacher-student consistency training with EMA improves detection on unlabeled roadside data.
    Section 3.3 assumes the standard SSOD paradigm without testing it against the supervised lower bound or alternative SSL methods.
  • domain assumption Freezing the ResNet-50 backbone parameters is harmless for the SSOD task.
    Section 3.4 fixes the backbone, which may limit the student's ability to adapt features to the unlabeled domain; the paper offers no ablation.
invented entities (4)
  • Dynamic pseudo-label assignment module
    purpose: Replaces IoU-based assignment with matching-cost-based reassignment of anchors
    Introduced in Section 3.3 with no algorithmic definition, no ablation, and no code.
  • Pseudo-feature alignment module
    purpose: Aligns classification and regression feature vectors across scales and locations
    Described qualitatively in Section 3.3; no equations or targeted experiments.
  • Pseudo-label refinement module
    purpose: Dynamically adjusts pseudo-box score thresholds to stabilize ground-truth counts
    Mentioned in Section 3.3; no thresholding rule is given.
  • Annotation alignment
    purpose: Maps synonymous textual descriptions to seven vehicle classes
    The clustering is done manually (Section 3.1) and is not compared to alternative taxonomies, so its benefit could be an artifact of the chosen label space.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Co-Learning: Towards Semi-Supervised Object Detection with Road-side Cameras." pith.science (2026). https://pith.science/paper/3EEP6F36

@misc{pith2026241119143,
  author       = {Pith},
  title        = {Pith review of: Co-Learning: Towards Semi-Supervised Object Detection with Road-side Cameras},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3EEP6F36}},
  note         = {Machine review of arXiv:2411.19143}
}
read the original abstract

Recently, deep learning has experienced rapid expansion, contributing significantly to the progress of supervised learning methodologies. However, acquiring labeled data in real-world settings can be costly, labor-intensive, and sometimes scarce. This challenge inhibits the extensive use of neural networks for practical tasks due to the impractical nature of labeling vast datasets for every individual application. To tackle this, semi-supervised learning (SSL) offers a promising solution by using both labeled and unlabeled data to train object detectors, potentially enhancing detection efficacy and reducing annotation costs. Nevertheless, SSL faces several challenges, including pseudo-target inconsistencies, disharmony between classification and regression tasks, and efficient use of abundant unlabeled data, especially on edge devices, such as roadside cameras. Thus, we developed a teacher-student-based SSL framework, Co-Learning, which employs mutual learning and annotation-alignment strategies to adeptly navigate these complexities and achieves comparable performance as fully-supervised solutions using 10\% labeled data.

Figures

Figures reproduced from arXiv: 2411.19143 by the authors.

Figure 1
Figure 1. Overview of the proposed Co-Learning framework: three modules to address the inconsistency in SSOD on computing-power-constrained devices, (𝑖) Pseudo-labels are determined by dynamic thresholds; (𝑖𝑖) Consistent pseudo-labels across teacher and student networks contribute to aligning and calibrating regression quality; (𝑖𝑖𝑖) Multi-head student networks reassign anchors based on matching cost. To achieve this, based o… view at source ↗
Figure 2
Figure 2. Qualititive Analysis of the proposed Co-Learning solution on the 7th AI CITY CHALLENGE Track 2 dataset. The base detector is Faster-RCNN with ResNet-50 as the feature extractor for all experiments to ensure a fair comparison. (DETs: Detected Results, GTs: Ground Truths) Methods Sedan Bus Pickup-Truck SUV Hatchback Van Truck mAP Oracle 26.2 76.6 31.8 17.6 0.6 29.5 0.5 36.1 w/o Annotation-Alignment 14.4 73.0 14.2 10.0… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 18 canonical work pages

  1. [1]

    Z. Zou, K. Chen, Z. Shi, Y. Guo, J. Ye, Object detection in 20 years: A survey, Proceedings of the IEEE 111 (2023) 257–276

    work page 2023

  2. [2]

    X. Qu, Q. Hu, S. Wang, Privacy-preserving model training architecture for intelligent edge computing, Computer Communications 162 (2020) 94–101

    work page 2020

  3. [3]

    Jiang, Z

    S. Jiang, Z. Lin, Y. Li, Y. Shu, Y. Liu, Flexible high-resolution object detection on edge devices with tunable latency, in: Proceedings of the 27th Annual International Conference on Mobile Computing and Networking, 2021, pp. 559–572

    work page 2021

  4. [4]

    Naphade, S

    M. Naphade, S. Wang, D. C. Anastasiu, Z. Tang, M.-C. Chang, Y. Yao, L. Zheng, M. S. Rahman, M. S. Arya, A. Sharma, Q. Feng, V. Ablavsky, S. Sclaroff, P. Chakraborty, S. Prajapati, A. Li, S. Li, K. Kunadharaju, S. Jiang, R. Chellappa, The 7th ai city challenge, in: The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023

    work page 2023

  5. [5]

    S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: Towards real-time object detection with region proposal networks, Advances in neural information processing systems 28 (2015)

    2015

  6. [6]

    Redmon, S

    J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: Unified, real-time object detection, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 779–788

    2016

  7. [7]

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, A. C. Berg, Ssd: Single shot multibox detector, in: Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14, Springer, 2016, pp. 21–37

    work page 2016

  8. [8]

    T.-Y. Lin, P. Goyal, R. Girshick, K. He, P. Dollár, Focal loss for dense object detection, in: Proceedings of the IEEE international conference on computer vision, 2017, pp. 2980–2988

    2017

Show all 26 references
  1. [9]

    Carion, F

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, S. Zagoruyko, End-to-end object detection with transformers, in: European conference on computer vision, Springer, 2020, pp. 213–229

    2020

  2. [10]

    Loganathan, G

    S. Loganathan, G. Kariyawasam, P. Sumathipala, Suspicious activity detection in surveil- lance footage, in: 2019 International Conference on Electrical and Computing technologies and applications (ICECTA), IEEE, 2019, pp. 1–4

    2019

  3. [11]

    Benbarrad, M

    T. Benbarrad, M. Salhaoui, S. B. Kenitar, M. Arioua, Intelligent machine vision model for defective product inspection based on machine learning, Journal of Sensor and Actuator Networks 10 (2021) 7

    2021

  4. [12]

    Jeong, S

    J. Jeong, S. Lee, J. Kim, N. Kwak, Consistency-based semi-supervised learning for object detection, Advances in neural information processing systems 32 (2019)

    2019

  5. [13]

    Y. Li, D. Huang, D. Qin, L. Wang, B. Gong, Improving object detection with selective self-supervised self-training, in: European Conference on Computer Vision, Springer, 2020, pp. 589–607

    2020

  6. [14]

    Liu, C.-Y

    Y.-C. Liu, C.-Y. Ma, Z. He, C.-W. Kuo, K. Chen, P. Zhang, B. Wu, Z. Kira, P. Vajda, Unbiased teacher for semi-supervised object detection, arXiv preprint arXiv:2102.09480 (2021)

    2021 arXiv

  7. [15]

    M. Xu, Z. Zhang, H. Hu, J. Wang, L. Wang, F. Wei, X. Bai, Z. Liu, End-to-end semi- supervised object detection with soft teacher, in: Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 3060–3069

    2021

  8. [16]

    B. Chen, P. Li, X. Chen, B. Wang, L. Zhang, X.-S. Hua, Dense learning based semi-supervised object detection, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 4815–4824

    2022

  9. [17]

    Zhang, X

    Y. Zhang, X. Yao, C. Liu, F. Chen, X. Song, T. Xing, R. Hu, H. Chai, P. Xu, G. Zhang, S4od: Semi-supervised learning for single-stage object detection, arXiv preprint arXiv:2204.04492 (2022)

    2022 arXiv

  10. [18]

    H. Zhou, Z. Ge, S. Liu, W. Mao, Z. Li, H. Yu, J. Sun, Dense teacher: Dense pseudo-labels for semi-supervised object detection, in: European Conference on Computer Vision, Springer, 2022, pp. 35–50

    2022

  11. [19]

    D. Xie, L. Liu, S. Zhang, J. Tian, A unified multi-modal structure for retrieving tracked ve- hicles through natural language descriptions, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 5419–5427

    2023

  12. [20]

    H. D.-A. Le, Q. Q.-V. Nguyen, D. T. Luu, T. T.-T. Chau, N. M. Chung, S. V.-U. Ha, Tracked- vehicle retrieval by natural language descriptions with multi-contextual adaptive knowl- edge, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023...

    2023

  13. [21]

    P. Sun, H. Kretzschmar, X. Dotiwalla, A. Chouard, V. Patnaik, P. Tsui, J. Guo, Y. Zhou, Y. Chai, B. Caine, et al., Scalability in perception for autonomous driving: Waymo open dataset, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020,...

    2020

  14. [22]

    T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Dollár, C. L. Zitnick, Microsoft coco: Common objects in context, in: Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, Springer, 2014, pp...

    2014

  15. [23]

    Palmer, D

    M. Palmer, D. Gildea, N. Xue, Semantic role labeling, Morgan & Claypool Publishers, 2011

    2011

  16. [24]

    Snoek, H

    J. Snoek, H. Larochelle, R. P. Adams, Practical bayesian optimization of machine learning algorithms, Advances in neural information processing systems 25 (2012)

    2012

  17. [25]

    maintainers, contributors, Torchvision: Pytorch’s computer vision library, https://github

    T. maintainers, contributors, Torchvision: Pytorch’s computer vision library, https://github. com/pytorch/vision, 2016

    2016

  18. [26]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Pro- ceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    2016

Pith tools

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