Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Endangered Alert: A Field-Validated Self-Training Scheme for Detecting and Protecting Threatened Wildlife on Roads and Roadsides

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

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

Pith's one-line read The paper claims that a roadside camera system can learn to spot a rare endangered bird without any human labelling, by cycling its own detections through a cloud auto-labeller and back into the edge model.

desk verdict Real deployment, real data, but the train/eval split ambiguity threatens the headline self-training trend; needs a leakage experiment before the claim is accepted. read the letter →

arxiv 2412.12222 v1 pith:YLT4NNQK submitted 2024-12-16 cs.CV

Kunming Li , Mao Shan , Stephany Berrio Perez , Katie Luo , Stewart Worrall This is my paper · ORCID
classification cs.CV
keywords self-trainingobjectdetectionrarewildlifecassowaryvision-languagemodelpseudo-labellingedgecomputingroadsideanimal
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 a roadside detection system for a rare and endangered bird, the cassowary, can improve itself over a five-month field deployment without any manual labelling. The system cycles data between an edge camera and a cloud server: the edge model selects the most informative frames, the cloud auto-labels them with a label-augmented vision-language detector, and the edge model is fine-tuned on those pseudo-labels. Per-frame detection of cassowaries within 100 metres rose from 4.2% to 78.5% across nine self-training iterations, and the full alert system caught 97% of cassowary events over 62 days. If correct, this means rare-species road alerts can be deployed in remote, low-bandwidth sites where hand-labelled data are unavailable.

What carries the argument

The load-bearing mechanism is Label-Augmentation Non-Maximum Suppression (LA-NMS), a pseudo-labelling step that feeds a vision-language detector both the target label and related labels (such as 'flightless bird' and 'black ratite bird' obtained from a lexical database), takes the maximum score over that expanded label set, and then applies non-maximum suppression to remove overlapping boxes. This widens the detector's coverage of a rare class it rarely saw during training, producing the auto-labels that drive the self-training loop. The complementary mechanism is selective data transfer: the edge device uploads only 1.5% of frames, prioritising true and false positive detections, so the cloud sees the most informative examples without saturating the link.

What would settle it

Recompute event-level recall with the event-triggering Bayesian aggregation layer disabled and a single fixed threshold on the raw edge-model detections; if recall drops to near the per-frame mTPR of 78.5% (and the 0.97 figure came largely from the aggregation), the system-level claim is not supported by the self-training loop.

Watch

Extended reading notes

Core claim

The central discovery is that self-training without manual annotation is sufficient to turn a synthetic-only detector with 4.2% mean true positive rate into a field detector with 78.5% mTPR at 0–100 m and 30.0% at 100–200 m, while keeping false positives around 0.37% of frames. The loop works by having the deployed edge model upload only 1.5% of recorded frames—chiefly its true and false positives—to the cloud, where Label-Augmentation Non-Maximum Suppression (LA-NMS) produces pseudo-labels that are then used to fine-tune the edge model. At the event level, the deployed system triggered on 194 true cassowary events and missed 6 over 62 days, yielding recall 0.97 and precision 0.77.

Load-bearing premise

The claim that the system catches 97% of cassowary events presumes that the under-described event-triggering pipeline, which aggregates Bayesian-filtered outputs from multiple cameras, is not itself tuned to the trial's ground truth; if that aggregation layer carries the recall, then the high event-level figure is not a clean measure of the self-training detector.

Editorial extensions

If this is right

  • Each fine-tuning round with auto-labelled field data raises near-range detection: the final model M240618 reaches 78.5% mTPR at 0–100 m, up from 4.2% for the synthetic-only initial model.
  • The approach holds false alarms roughly constant (FPR about 0.37%) while improving recall, and produced fewer than one false alert per day during the on-road trial.
  • The identical scheme transfers to thermal imagery, indicating that the self-training loop adapts across camera modalities and lighting conditions.
  • The deployed system missed only 6 of 200 cassowary sighting events over 62 days, a level of recall that would be necessary for driver-alert credibility.

Reading between the lines

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

  • The label-augmentation trick should transfer to other rare-object detection tasks where the target class has semantically close relatives in a lexical database; the main risk is that poorly chosen augmentations (as the paper notes for sheep and cow) increase false positives.
  • The 1.5% selective upload suggests a general bandwidth-saving principle for edge self-training: a detector's own errors are the most informative training signal, so upload policies should prioritise high-uncertainty frames rather than random sampling.
  • The reported event-level recall of 0.97 is very likely inflated relative to per-frame performance because of the separate Bayesian aggregation layer; a clean evaluation of the self-training contribution would report both metrics with that layer ablated.
  • The Stage 1 synthetic-only model's 4.2% mTPR suggests that synthetic pre-training alone is far too weak for safety-critical deployment, and that the value of the pipeline comes from the real auto-labelled data loop rather than the synthesis itself.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper describes a self-training pipeline for detecting rare animals (cassowaries) on roads and roadsides, combining an edge-deployed YOLOv8 detector with a cloud-based vision-language-model pipeline that pseudo-labels selectively transferred field data. Stage 1 generates synthetic training data from web images, and Stage 2 iteratively fine-tunes the field model on auto-labelled field data. The authors validate LA-NMS on the VOC benchmark and report results from a five-month deployment, including a per-frame mTPR improvement from 4.2% to 78.5% across successive models and an event-level precision/recall of 0.77/0.97 during a 62-day on-road trial.

Significance. If the central claim is fully supported, the work is valuable for roadside animal detection in resource-limited environments because it demonstrates a real deployment, a manually verified evaluation set, and a concrete mechanism for reducing manual labelling effort. The paper also has concrete strengths: the five-month field trial is a genuine contribution, the manual verification of 259 events provides a useful ground-truth resource, the source code is released, and the LA-NMS idea is simple and transparently evaluated on the VOC benchmark. The main gap is that the evidence for the self-training trend itself is not yet controlled enough to separate improved data quantity from the iterative self-training mechanism, and the system-level recall depends on an under-specified aggregation stage.

major comments (5)
  1. [V-A, V-C.1, Table II] The paper does not specify whether the fixed evaluation set is disjoint from the dynamically updated training subset, either temporally or by sighting case. Section V-A states that the field training subset is maintained at 4,000-6,000 images while the evaluation set is a fixed collection of 4,577 cassowary images from 38 sighting cases, and Section V-C.1 only says the models were trained and validated with distinct datasets. If images or sighting cases from the evaluation period appear in the training subset of later models, the monotone mTPR increase in Table II would be expected even without any self-training benefit. The authors should state the split criterion and, ideally, re-evaluate the models on a hold-out set that is provably disjoint from all training iterations.
  2. [VI-B, Table II] The main evidence for self-training improvement has no controlled baseline: there is no fully supervised model trained on the same data, no comparison with a standard semi-supervised object detection method, no error bars or variance over repeats, and later models are trained on progressively more data. The observed trend conflates the self-training mechanism with simply using more training images. An ablation that trains a single model on the pooled field data available at the final date, without iterative pseudo-labelling, would directly test whether the iterative loop adds value beyond data quantity.
  3. [III-E, VI-A] Pseudo-label quality is not measured. The labels used for fine-tuning are produced by the OWL-VIT/LA-NMS pipeline, and the evaluation of that same pipeline in Section VI-A uses the same model family and the same field evaluation set; this creates a real risk that self-training entrenches systematic VLM errors. The authors should report pseudo-label precision/recall against a manually labelled subset of the transferred field images, and show how pseudo-label noise propagates through the fine-tuning iterations.
  4. [VI-B] The event-level recall of 0.97 is attributed to an 'event-triggering pipeline' that aggregates Bayesian-filtered results from multiple Wfield instances across cameras, but the Bayesian filter, its parameters, and whether it was tuned on the trial data are not described. As written, the high event recall is a system-level result that cannot be assigned to the self-training detector. The authors should document the aggregation method, or report per-camera, per-frame detector metrics as the primary claim.
  5. [VI-B, Table II] The comparison across models uses a post hoc FPR cut-off of 0.4%, while all reported FPR values are already clustered around 0.35-0.39%. This operation-point selection can flatter specific models and does not convey the uncertainty or the precision-recall trade-off. The authors should justify the cut-off, report ROC curves with confidence intervals, and give detector precision/recall at matched operating points.
minor comments (5)
  1. [VI-B] In the paragraph after Table II, the text refers to 'M240628' when the corresponding table and context consistently use 'M240618'; this should be corrected.
  2. [II-A, IV] There are typographical errors: 'psudo groundtruth mining' in Section II-A and 'therfore' in Section IV should be corrected.
  3. [V-B] The mTPR definition says it averages TPR over images in the test dataset, but Table II reports averages over sighting cases in each range group; the unit of aggregation should be stated consistently.
  4. [VI-A, Fig. 5] The y-axis label in Fig. 5 is missing or illegible in the provided version; adding an explicit 'mAP' label would improve readability.
  5. [References] Reference [18] is malformed ('Q. Y . C-mil' appears as an author), and several references lack venue or year formatting consistency; a final proofread of the bibliography is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No definitional circularity: the claimed self-training improvement is evaluated with a manually verified field evaluation set and an external VOC benchmark, not with the system's own pseudo-labels.

full rationale

The central claim is that the iterative self-training pipeline improves Wfield without manual labelling. The evidence for this is Tab. II, where ten chronologically trained models are evaluated on a fixed field evaluation dataset containing 4,577 manually labelled cassowary images, and Fig. 4, where LA-NMS is validated on the public VOC dataset. Neither the training objective nor the evaluation metric is defined in terms of the claimed outcome: Wfield is trained on pseudo-labels produced by an external VLM (OWL-VIT) and SAM, while the evaluation uses human-verified labels. The paper explicitly states that 'manual data inspection in this work is solely for performance evaluation,' which blocks the most direct circularity. The LA-NMS component is itself assessed against VOC and field data before being used as a label generator, so its contribution is not assumed by construction. There are no load-bearing self-citations, and no 'uniqueness' or 'ansatz' is imported from the authors' prior work. The main residual concern is experimental rather than circular: the paper does not specify a temporal or case-level disjointness guarantee between the dynamically updated training subset and the 38-sighting evaluation set, so later models such as M240618 could in principle have seen evaluation-period frames during training. If that were the case, the monotone mTPR trend in Tab. II would partly reflect training-set accuracy rather than generalization. However, that would be an evaluation-contamination or experimental-design issue, not a case where the prediction is equivalent to the input by definition. The under-specified Bayesian event-aggregation pipeline affects the system-level recall of 0.97 but is separate from the self-training claim and is not circular. Because no specific reduction from output to input can be exhibited from the text, the appropriate finding is no significant circularity.

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

The paper's central claim relies on the VLM's zero-shot competence, the validity of synthetic training data, the exhaustiveness of manual evaluation, and the unstated event-triggering filter. None of these are fitted to the target metric, but they are substantial domain assumptions.

free parameters (4)
  • Pseudo-label confidence threshold = not stated (described only as 'higher threshold')
    Controls which VLM detections become training labels; the paper does not report the value or a sensitivity study (Section III-E, VI-A).
  • FPR cut-off threshold = 0.4%
    Used to compare trained models at a matched false positive rate; selection of this cut-off is post hoc and affects the mTPR rankings in Table II (Section VI-B).
  • Data transfer ratio = 1.5%
    The fraction of recorded field data uploaded to the cloud; a design choice that affects which examples enter the training set (Section III-E).
  • Fine-tuning layer fraction = 50%
    The fraction of YOLOv8 layers unfrozen during fine-tuning; chosen without reported ablation (Section III-F2).
assumptions (4)
  • domain assumption OWL-ViT with text prompts can produce usable pseudo-labels for cassowaries after WordNet label augmentation.
    The entire auto-labelling pipeline depends on the VLM's zero-shot detections being accurate enough; validated only indirectly via VOC and field mAP comparisons, not against manual labels on the training set (Section IV, VI-A).
  • domain assumption Synthetic images created by cut-paste with Gaussian blending are a sufficient proxy for real field images for initial training.
    Stage 1 trains the initial model exclusively on synthetic data; if the sim-to-real gap is large, the initial model would be poor, though Stage 2 real data could compensate (Section III-D).
  • domain assumption Manual inspection of recorded videos is exhaustive enough to identify all cassowary appearances, including those the system missed.
    The reported FN count and recall 0.97 rest on the completeness of this manual review (Section VI-B).
  • domain assumption The Bayesian-filter event-triggering pipeline correctly aggregates per-camera detections without introducing its own tuned parameters.
    System-level recall (0.97) is computed from events generated by this pipeline, whose details are not given (Section VI-B).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Endangered Alert: A Field-Validated Self-Training Scheme for Detecting and Protecting Threatened Wildlife on Roads and Roadsides." pith.science (2026). https://pith.science/paper/YLT4NNQK

@misc{pith2026241212222,
  author       = {Pith},
  title        = {Pith review of: Endangered Alert: A Field-Validated Self-Training Scheme for Detecting and Protecting Threatened Wildlife on Roads and Roadsides},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YLT4NNQK}},
  note         = {Machine review of arXiv:2412.12222}
}
read the original abstract

Traffic accidents are a global safety concern, resulting in numerous fatalities each year. A considerable number of these deaths are caused by animal-vehicle collisions (AVCs), which not only endanger human lives but also present serious risks to animal populations. This paper presents an innovative self-training methodology aimed at detecting rare animals, such as the cassowary in Australia, whose survival is threatened by road accidents. The proposed method addresses critical real-world challenges, including acquiring and labelling sensor data for rare animal species in resource-limited environments. It achieves this by leveraging cloud and edge computing, and automatic data labelling to improve the detection performance of the field-deployed model iteratively. Our approach introduces Label-Augmentation Non-Maximum Suppression (LA-NMS), which incorporates a vision-language model (VLM) to enable automated data labelling. During a five-month deployment, we confirmed the method's robustness and effectiveness, resulting in improved object detection accuracy and increased prediction confidence. The source code is available: https://github.com/acfr/CassDetect

Figures

Figures reproduced from arXiv: 2412.12222 by the authors.

Figure 1
Figure 1. Deployed system identifying a cassowary crossing a busy road. Upon detecting a cassowary, as illustrated in (a), the system alerts approaching vehicles through a variable message sign (VMS) to help prevent potential collisions, as shown in (b). (Message design credit: Ioni Lewis, Queensland University of Technology) the scalability and adaptability of these technologies across diverse geographic and climatic conditi… view at source ↗
Figure 2
Figure 2. Overview of the proposed self-training ML scheme for roadside animal detection. Initially, the cloud-based model Wcloud synthesises images of cassowaries and generates the pseudo-labels using the web-sourced cassowary images and field background images. These images are used to train the initial field detection model for deployment on the edge device. In the deployment environment, this field model Wfield processes … view at source ↗
Figure 3
Figure 3. Workflow of the proposed LA-NMS. The approach begins by taking an input animal label and generating multiple related class labels. These augmented labels and input images are encoded using a text transformer encoder and a vision transformer encoder, respectively, within a VLM, i.e., OWL-VIT in this work. This process facilitates the identification of detection candidates in the input images based on the augmented la… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of mAP for various object classes from the VOC dataset, analysed with and without LA-NMS. ‘Label Aug.’ in the legend refers to LA-NMS [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: illustrates the performance comparison between the VLM with and without LA-NMS enabled across various confidence threshold settings on the field evaluation dataset. Calibration of the confidence threshold is critical in practical applications to ensure pseudo-labels’ a…
Figure 6
Figure 6. Figure 6: ROC curves of the trained models on RGB and thermal images for cassowary crossing scenes. Method Field Data Web-Sourced Dataset w/o LA-NMS 0.683 0.472 w/ LA-NMS 0.795 0.529 TABLE I: mAP of the VLM with and without LA-NMS enabled for field data and web-sourced data. Sta…
Figure 7
Figure 7. Figure 7: Distribution of cassowary detection events over time. There were 166 cassowary sighting cases in May 2024 before a steep drop in June 2024. Each event has been checked against the ground truth data to classify them as TP, FN, and FP detection events. Despite the system…
Figure 8
Figure 8. Figure 8: Demonstrating the system’s detection of a cassowary in thermal imagery from the field trial. The experiment evaluates the adaptability of the pro￾posed method to the thermal domain, which is essential for validating its robustness under inadequate lighting condi￾tions …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Animal Interaction with Autonomous Mobility Systems: Designing for Multi-Species Coexistence

    cs.HC 2025-07 conditional novelty 5.0 of 10

    Animal encounters with autonomous vehicles, delivery robots, and drones cluster into five concern areas, physical impact, behavioral effects, accessibility, ethics and regulation, and urban disturbance, with design an...

Reference graph

Works this paper leans on

52 extracted references · 47 canonical work pages · cited by 1 Pith paper

  1. [1]

    World Health Organization, 2019, accessed on October 24, 2024

    World Health Organization, Global status report on road safety 2018 . World Health Organization, 2019, accessed on October 24, 2024

    work page 2018

  2. [2]

    Wildlife road safety report reveals dangers of animal collisions,

    Insurance Australia Group, “Wildlife road safety report reveals dangers of animal collisions,” https://www.iag.com.au/newsroom/community/ wildlife-road-safety-report-reveals-dangers-animal-collisions, accessed on October 24, 2024

    work page 2024

  3. [3]

    (2023) Wildlife detection

    Navtech Radar. (2023) Wildlife detection. Accessed: 2023-9-12. [On- line]. Available: https://navtechradar.com/explore/wildlife-detection/

    work page 2023

  4. [4]

    Evaluation of an animal warning system effectiveness phase two,

    M. A. Sharafsaleh, M. Huijser, C. Nowakowski, M. C. Greenwood, L. Hayden, J. Felder, and M. Wang, “Evaluation of an animal warning system effectiveness phase two,” Tech. Rep., 2012

    work page 2012

  5. [5]

    Evaluation of a buried cable roadside an- imal detection system

    C. Druta, A. S. Alden et al., “Evaluation of a buried cable roadside an- imal detection system.” Virginia Center for Transportation Innovation and Research, Tech. Rep., 2015

    work page 2015

  6. [6]

    Preventing animal-vehicle crashes using a smart detection technology and warning system,

    C. Druta and A. S. Alden, “Preventing animal-vehicle crashes using a smart detection technology and warning system,” Transportation research record, 2020

    work page 2020

  7. [7]

    Intelligent system utilizing hog and cnn for thermal image-based detection of wild animals in nocturnal periods for vehicle safety,

    Y . Munian, A. Martinez-Molina, D. Miserlis, H. Hernandez, and M. Alamaniotis, “Intelligent system utilizing hog and cnn for thermal image-based detection of wild animals in nocturnal periods for vehicle safety,”Applied Artificial Intelligence, vol. 36, no. 1, p. 2031825, 2022

    work page 2022

  8. [8]

    Multiple instance detection network with online instance classifier refinement,

    P. Tang, X. Wang, X. Bai, and W. Liu, “Multiple instance detection network with online instance classifier refinement,” in Proc. CVPR , 2017

    work page 2017

Show all 52 references
  1. [9]

    Object instance mining for weakly supervised object detection,

    C. Lin, S. Wang, D. Xu, Y . Lu, and W. Zhang, “Object instance mining for weakly supervised object detection,” in Proc. the AAAI Conference on Artificial Intelligence , 2020

    2020

  2. [10]

    Instance-aware, context-focused, and memory-efficient weakly supervised object detection,

    Z. Ren, Z. Yu, X. Yang, M.-Y . Liu, Y . J. Lee, A. G. Schwing, and J. Kautz, “Instance-aware, context-focused, and memory-efficient weakly supervised object detection,” in Proc. CVPR, 2020

    2020

  3. [11]

    Pcl: Proposal cluster learning for weakly supervised object detection,

    P. Tang, X. Wang, S. Bai, W. Shen, X. Bai, W. Liu, and A. Yuille, “Pcl: Proposal cluster learning for weakly supervised object detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2018

    2018

  4. [12]

    Instance mining with class feature banks for weakly supervised object detection,

    Y . Yin, J. Deng, W. Zhou, and H. Li, “Instance mining with class feature banks for weakly supervised object detection,” in Proc. the AAAI Conference on Artificial Intelligence , 2021

    2021

  5. [13]

    Comprehensive attention self-distillation for weakly-supervised object detection,

    Z. Huang, Y . Zou, B. Kumar, and D. Huang, “Comprehensive attention self-distillation for weakly-supervised object detection,” Advances in neural information processing systems , 2020

    2020

  6. [14]

    Cat: Weakly supervised object detection with category transfer,

    T. Cao, L. Du, X. Zhang, S. Chen, Y . Zhang, and Y .-F. Wang, “Cat: Weakly supervised object detection with category transfer,” in Proc. ICCV, 2021

    2021

  7. [15]

    Boost- ing weakly supervised object detection via learning bounding box adjusters,

    B. Dong, Z. Huang, Y . Guo, Q. Wang, Z. Niu, and W. Zuo, “Boost- ing weakly supervised object detection via learning bounding box adjusters,” in Proc. ICCV, 2021

    2021

  8. [16]

    Boosting weakly super- vised object detection with progressive knowledge transfer,

    Y . Zhong, J. Wang, J. Peng, and L. Zhang, “Boosting weakly super- vised object detection with progressive knowledge transfer,” in Proc. ECCV. Springer, 2020

    2020

  9. [17]

    Weakly supervised deep detection net- works,

    H. Bilen and A. Vedaldi, “Weakly supervised deep detection net- works,” in Proc. CVPR, 2016

    2016

  10. [18]

    Continuation multiple instance learning for weakly supervised object detection. 2019 ieee,

    F. Wan, C. Liu, W. Ke, X. Ji, J. Jiao, and Q. Y . C-mil, “Continuation multiple instance learning for weakly supervised object detection. 2019 ieee,” in Proc. CVPR, 2019

    2019

  11. [19]

    Towards human-machine cooperation: Self-supervised sample mining for object detection,

    K. Wang, X. Yan, D. Zhang, L. Zhang, and L. Lin, “Towards human-machine cooperation: Self-supervised sample mining for object detection,” in Proc. CVPR, 2018

    2018

  12. [20]

    Consistency-based semi- supervised learning for object detection,

    J. Jeong, S. Lee, J. Kim, and N. Kwak, “Consistency-based semi- supervised learning for object detection,” Advances in neural infor- mation processing systems , 2019

    2019

  13. [21]

    A simple semi-supervised learning framework for object detection,

    K. Sohn, Z. Zhang, C.-L. Li, H. Zhang, C.-Y . Lee, and T. Pfister, “A simple semi-supervised learning framework for object detection,” arXiv preprint arXiv:2005.04757 , 2020

    2005 arXiv

  14. [22]

    Unbiased teacher for semi-supervised object detection,

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

    2021 arXiv

  15. [23]

    Interactive self- training with mean teachers for semi-supervised object detection,

    Q. Yang, X. Wei, B. Wang, X.-S. Hua, and L. Zhang, “Interactive self- training with mean teachers for semi-supervised object detection,” in Proc. CVPR, 2021

    2021

  16. [24]

    Watch and learn: Semi- supervised learning for object detectors from video,

    I. Misra, A. Shrivastava, and M. Hebert, “Watch and learn: Semi- supervised learning for object detectors from video,” in Proc. CVPR, 2015

    2015

  17. [25]

    Large scale semi-supervised object detection using visual and seman- tic knowledge transfer,

    Y . Tang, J. Wang, B. Gao, E. Dellandr ´ea, R. Gaizauskas, and L. Chen, “Large scale semi-supervised object detection using visual and seman- tic knowledge transfer,” in Proc. CVPR, 2016

    2016

  18. [26]

    Automatic adaptation of object detectors to new domains using self-training,

    A. RoyChowdhury, P. Chakrabarty, A. Singh, S. Jin, H. Jiang, L. Cao, and E. Learned-Miller, “Automatic adaptation of object detectors to new domains using self-training,” in Proc. CVPR, 2019

    2019

  19. [27]

    Semi-supervised self-training of object detection models,

    C. Rosenberg, M. Hebert, and H. Schneiderman, “Semi-supervised self-training of object detection models,” 2005

    2005

  20. [28]

    Grounding dino 1.5: Advance the

    T. Ren, Q. Jiang, S. Liu, Z. Zeng, W. Liu, H. Gao, H. Huang, Z. Ma, X. Jiang, Y . Chenet al., “Grounding dino 1.5: Advance the” edge” of open-set object detection,” arXiv preprint arXiv:2405.10300 , 2024

    2024 arXiv

  21. [29]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection,

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, C. Li, J. Yang, H. Su, J. Zhu et al. , “Grounding dino: Marrying dino with grounded pre-training for open-set object detection,” arXiv preprint arXiv:2303.05499, 2023

    2023 arXiv

  22. [30]

    Learning transferable visual models from natural language supervision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark et al., “Learning transferable visual models from natural language supervision,” in Proc. Interna- tional conference on machine learning , 2021

    2021

  23. [31]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo et al. , “Segment anything,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023

    2023

  24. [32]

    Scaling open-vocabulary object detection,

    M. Minderer, A. Gritsenko, and N. Houlsby, “Scaling open-vocabulary object detection,” Advances in Neural Information Processing Systems, 2024

    2024

  25. [33]

    Self-train: Self-supervised on-device training for post-deployment adaptation,

    J. Liu, X. Yu, and T. Rosing, “Self-train: Self-supervised on-device training for post-deployment adaptation,” in Proc. IEEE International Conference on Smart Internet of Things (SmartIoT) , 2022

    2022

  26. [34]

    Train++: An incremental ml model training algorithm to create self-learning iot devices,

    B. Sudharsan, P. Yadav, J. G. Breslin, and M. I. Ali, “Train++: An incremental ml model training algorithm to create self-learning iot devices,” in Proc. SmartWorld/SCALCOM/UIC/ATC/IOP/SCI, 2021

    2021

  27. [35]

    List: Lite prompted self-training makes parameter-efficient few-shot learners,

    Y . Wang, S. Mukherjee, X. Liu, J. Gao, A. H. Awadallah, and J. Gao, “List: Lite prompted self-training makes parameter-efficient few-shot learners,” arXiv preprint arXiv:2110.06274 , 2021

    2021 arXiv

  28. [36]

    Tinytl: Reduce memory, not parameters for efficient on-device learning,

    H. Cai, C. Gan, L. Zhu, and S. Han, “Tinytl: Reduce memory, not parameters for efficient on-device learning,” Advances in Neural Information Processing Systems , 2020

    2020

  29. [37]

    Rep-net: Efficient on-device learning via feature reprogramming,

    L. Yang, A. S. Rakin, and D. Fan, “Rep-net: Efficient on-device learning via feature reprogramming,” in Proc. CVPR, 2022

    2022

  30. [38]

    Mobiletl: on-device transfer learning with inverted residual blocks,

    H.-Y . Chiang, N. Frumkin, F. Liang, and D. Marculescu, “Mobiletl: on-device transfer learning with inverted residual blocks,” in Proc. the AAAI Conference on Artificial Intelligence , 2023

    2023

  31. [39]

    A joint energy and latency framework for transfer learning over 5g industrial edge networks,

    B. Yang, O. Fagbohungbe, X. Cao, C. Yuen, L. Qian, D. Niyato, and Y . Zhang, “A joint energy and latency framework for transfer learning over 5g industrial edge networks,” IEEE Transactions on Industrial Informatics, 2021

    2021

  32. [40]

    Accelerating on-device dnn training workloads via runtime convergence monitor,

    S. Choi, J. Shin, and L.-S. Kim, “Accelerating on-device dnn training workloads via runtime convergence monitor,” IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems , 2022

    2022

  33. [41]

    Mobileda: Toward edge-domain adaptation,

    J. Yang, H. Zou, S. Cao, Z. Chen, and L. Xie, “Mobileda: Toward edge-domain adaptation,” IEEE Internet of Things Journal , 2020

    2020

  34. [42]

    Communication-efficient learning of deep networks from decentral- ized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentral- ized data,” in Artificial intelligence and statistics , 2017

    2017

  35. [43]

    Federated learning in edge computing: a systematic survey,

    H. G. Abreha, M. Hayajneh, and M. A. Serhani, “Federated learning in edge computing: a systematic survey,” Sensors, 2022

    2022

  36. [44]

    A novel federated edge learning ap- proach for detecting cyberattacks in iot infrastructures,

    S. Abbas, A. Al Hejaili, G. A. Sampedro, M. Abisado, A. Almadhor, T. Shahzad, and K. Ouahada, “A novel federated edge learning ap- proach for detecting cyberattacks in iot infrastructures,” IEEE Access, 2023

    2023

  37. [45]

    Federated learning for internet of things,

    T. Zhang, C. He, T. Ma, L. Gao, M. Ma, and S. Avestimehr, “Federated learning for internet of things,” in Proc. the 19th ACM Conference on Embedded Networked Sensor Systems , 2021

    2021

  38. [46]

    Yolov8: A novel object detection algorithm with enhanced performance and robustness,

    R. Varghese and S. M., “Yolov8: A novel object detection algorithm with enhanced performance and robustness,” in 2024 International Conference on Advances in Data Engineering and Intelligent Com- puting Systems (ADICS) , 2024, pp. 1–6

    2024

  39. [47]

    Wordnet,

    C. Fellbaum, “Wordnet,” in Theory and applications of ontology: computer applications, 2010

    2010

  40. [48]

    Beyond generation: Harnessing text to image models for object detection and segmentation,

    Y . Ge, J. Xu, B. N. Zhao, N. Joshi, L. Itti, and V . Vineet, “Beyond generation: Harnessing text to image models for object detection and segmentation,” arXiv preprint arXiv:2309.05956 , 2023

    2023 arXiv

  41. [49]

    Cut, paste and learn: Surpris- ingly easy synthesis for instance detection,

    D. Dwibedi, I. Misra, and M. Hebert, “Cut, paste and learn: Surpris- ingly easy synthesis for instance detection,” in Proc. ICCV, 2017

    2017

  42. [50]

    Simple copy-paste is a strong data augmentation method for instance segmentation,

    G. Ghiasi, Y . Cui, A. Srinivas, R. Qian, T.-Y . Lin, E. D. Cubuk, Q. V . Le, and B. Zoph, “Simple copy-paste is a strong data augmentation method for instance segmentation,” in Proc. CVPR, 2021

    2021

  43. [51]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Proc. ECCV, 2014

    2014

  44. [52]

    Cassowary casques act as thermal windows,

    D. L. Eastick, G. J. Tattersall, S. J. Watson, J. A. Lesku, and K. A. Robert, “Cassowary casques act as thermal windows,” Scientific Reports, vol. 9, no. 1, p. 1966, 2019

    1966

Pith tools

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