The DOI matches its Crossref record; 153 of the 153 references it cites were found in a registry or an index; 37 of the 184 sources its claims cite were checked for support.
A Systematic Review of Drone Based Road Traffic Monitoring System
Open accessAbstract
Drone deployment has become crucial in a variety of applications, including solutions to traffic issues in metropolitan areas and highways. On the other hand, data collected via drones suffers from several problems, including a wide range of object scales, angle variations, truncation, and occlusion. To process and manipulate visual data from the drones, a variety of image processing algorithms have been employed, each with a distinct aim. Additionally, recent breakthroughs in the field of Artificial Intelligence, particularly deep learning, have attracted broad interest and are being applied to many domains in the framework of smart cities, including road traffic monitoring. The purpose of this study is to conduct a systematic review of drone-based traffic monitoring systems from a deep learning perspective. This work focuses on vehicle detection, tracking, and counting, since they are fundamental building blocks towards founding solutions for traffic congestion, flow rate and vehicle speed estimation. Additionally, drone-based datasets are examined, which face issues and problems caused by the diversity of features inherent of drone devices. The review analysis presented in this work summarizes the literature solutions provided and deployed so far and discusses future research trends in establishing a comprehensive traffic monitoring system in support of the development of smart cities.
Overview
CoChat Summary
TL;DR
This systematic review surveys deep-learning approaches for drone-based road traffic monitoring, focusing on vehicle detection, tracking, and counting, plus aerial datasets and evaluation metrics. It finds that custom datasets often yield >80% performance while public datasets such as VisDrone and DOTA show lower results, and identifies the lack of a large-scale diverse drone traffic dataset as a key open issue.
Key findings
- The review screened 37 journals, 52 conference papers, and 9 other items, including 19 journals and 13 conference papers, for the publication period 2015–2021; the shortlisted articles total 32 across detection, counting, and tracking (Tables 1–2).
- Most systems use one- or two-stage detectors, with Faster R-CNN and YOLOv3 most widely employed; custom datasets often achieve >80% performance, while public datasets such as VisDrone and DOTA yield lower performance.
- Counting accuracy exceeds 90% with custom datasets (e.g., 93.7% counting-by-tracking [3], 97.58% daytime lane-based counting [76]), and tracking performance is >80% in the majority of studies (e.g., MOSSE tracker MOTA 90.91%, MOTP 92.11% [5]).
- Nighttime detection/counting is lower than daytime (e.g., 84.6% night vs 87.9% day detection [76]; 85.26% night vs 97.58% day counting [76]), and a drone-based large-scale diverse aerial dataset comparable to MS-COCO is still missing.
Methods
- Systematic review with research questions on datasets/metrics (RQ1), DL frameworks and pre-processing for detection/tracking/counting (RQ2), and effectiveness/challenges/future trends (RQ3); two-step screening by title/abstract then full-text reading, plus reference scanning.
- Inclusion: English-language publications from 2015–2021; journals, conferences, and technical reports were screened, with final inclusion of 19 journal and 13 conference articles (Table 1).
- Analysis organizes literature into image-based detection, video-based detection, on-board monitoring, pre-processing, counting, and tracking, with comparison tables of frameworks, datasets, metrics, and outcomes.
Limitations
- Paper-stated: exact UAV altitude is not always reported; public datasets yield lower performance; custom datasets often use limited object classes (up to four target classes in many systems); nighttime performance is lower; tracking outcomes are not directly comparable because systems use different datasets and class counts.
- Paper-stated: a drone-based large-scale diverse aerial dataset comparable to MS-COCO is still lacking; on-board processing is computationally costly in power consumption, flight duration, and battery life; privacy concerns can force higher altitudes or side views, reducing performance.
- Evident from text: the review is restricted to 2015–2021 English-language literature and focuses solely on detection, tracking, and counting, excluding congestion analysis, flow rate, and speed estimation despite noting them as traffic monitoring tasks.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
15 Valid · 21 Not valid · 122 Not checked · 69 Own results and interpretations
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Results increased over base YOLOv3 [25], Faster R-CNN [84] and SSD [82] frameworks.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > A. TRAFFIC OBJECT DETECTION 1) Image-based Detection
What the sources say
- Supportedcites [25] Redmon, 2018, YOLOv3: An incremental improvement source
The title and Abstract (p. 1) identify the cited work as YOLOv3. In the citing sentence, [25] identifies the base YOLOv3 framework used as a comparator; the reported improvement is the citing authors’ result, not a result attributed to this source. (LLM, conf=0.92)
- Not checked — no full text of the sourcecites [84] source
- Disputedcites [82] sourceExperimental — accuracy not yet measured
The citation is attached to “base YOLOv3,” but the Conclusion (p. 8) identifies this paper’s systems as YOLOv2 and YOLO9000, not YOLOv3. Its comparisons with Faster R-CNN and SSD in “Better” (p. 4) likewise concern YOLOv2; they do not report ISPDM results. (LLM, conf=0.88)
Checked by a reasoning model — may be wrong.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“For this reason, [69] proposed a vehicle classification system for drone imagery that combines transfer learning features from the Inception-ResNet-v2 model [79] with hand-made features from HoG [22], Local Binary Patters (LBP) [80] and Bag-of-Visual-Words (BoVW) [81].”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > A. TRAFFIC OBJECT DETECTION 1) Image-based Detection
What the sources say
- Disputedcites [69] sourceExperimental — accuracy not yet measured
Our copy of Li, Li and Pan (2020) proposes a vehicle DETECTION algorithm (SSP-SSD), not a classification system fusing transfer-learning and hand-made features: the Abstract and Introduction list the contributions as adaptive image cropping, scale-specific prediction on an SSD detector, and Outlier-Aware NMS, and Section B (Scale-Specific Prediction) and Section E (Performance Evaluation) state 'The detection network is built on SSD... We utilize Resnet101 [42] as backbone for feature extraction' — Inception-ResNet-v2 is never mentioned anywhere in the paper. HoG and LBP appear only in Section II.A (Traditional Vehicle Detection) as descriptors used by OTHER works ([13], [26], [27], [30]), where the authors conclude that 'it is difficult for handcrafted descriptors to completely distinguish vehicles from the background'; Bag-of-Words appears only in reference [28] (Zhou et al.). Thus the cited sentence attributes to this source a method and feature combination it explicitly rejects and does not use. (LLM, conf=0.85)
Checked by a reasoning model — may be wrong.
- Not checked — no full text of the sourcecites [79] source
- Supportedcites [22] source
The citation [22] is attached to HoG as a feature used in the citing sentence, not to the vehicle-classification system attributed to [69]. The source paper’s Abstract identifies and studies Histograms of Oriented Gradient (HOG) descriptors, so it supports that identification. (LLM, conf=0.93)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
Every source the paper cites for this claim was checked and supports it.
In the paper
“Likewise, UAV Detection and Tracking (UAVDT) dataset [45] provides information for video object detection, SOT, and MOT tasks with vehicles as the target objects.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > IV. AERIAL DATASETS AND EVALUATING METRICS (RQ1) A. DATASETS
What the sources say
- Supportedcites [45] source
The Introduction (p. 2) identifies UAVDT as a benchmark for object detection, single-object tracking (SOT), and multiple-object tracking (MOT), and states that its objects of interest are vehicles. The Conclusion (p. 14) confirms that the dataset consists of UAV-captured videos for those three tasks. (LLM, conf=0.94)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, since drone images have aerial views with occlusions, varying sizes and orientations, employing DL models pre-trained on datasets such as Imagenet [57] and MS-COCO [78] datasets might not provide high performance.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > A. TRAFFIC OBJECT DETECTION 1) Image-based Detection
What the sources say
- Partly supportedcites [57] sourceExperimental — accuracy not yet measured
Our copy supports only the dataset/background half of the sentence: Section 3 (Bounding Box Formats) identifies COCO, PASCAL VOC and Open Images as the popular competition datasets used to train and evaluate detectors, Table 2 / 'Benchmark Dataset Metrics' lists COCO- and ImageNet-evaluated CNN detectors, and Sections 6.5 and 6.10 note that size-aware AP_S/AP_M/AP_L metrics should be used 'in datasets where the objects appear to have relatively different sizes'. Nothing in our copy addresses drone/UAV/aerial imagery, occlusion in aerial views, or the proposition that models pre-trained on ImageNet or MS-COCO fail to deliver high performance there (searches: drone, unmanned aerial vehicle, aerial, pre-trained/transfer learning, ImageNet-COCO performance). The causal claim about pre-training limits on drone data is therefore not attributable to this source. (LLM, conf=0.65)
- Could not verifycites [78] sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) identifies ImageNet as an image-recognition benchmark, and the Experimental Results report performance on that benchmark. Our copy does not address drone imagery, its occlusions or varied viewpoints, or whether ImageNet- or MS-COCO-pre-trained models perform poorly on drone images. Searches for those terms and for aerial-image transfer learning found no supporting passage. (LLM, conf=0.76)
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Nonetheless, this strategy provides scarce performance when compared with SSD [82], You Only Look Once (YOLO)9000 [83], and YOLOv3 [25] frameworks.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > A. TRAFFIC OBJECT DETECTION 1) Image-based Detection
What the sources say
- Supportedcites [82] source
The citation is attached to YOLO9000 as one of the named comparison frameworks. The paper introduces YOLO9000 as a real-time object detector (p. 1), so it supports that identification; the citing authors’ assessment of SA-CF performance is not attributed specifically to this source. (LLM, conf=0.94)
- Could not verifycites [83] sourceExperimental — accuracy not yet measured
The experiments on PASCAL VOC (pp. 8–10) and MS COCO (p. 11) report Faster R-CNN results and comparisons with other methods, but not an SA-CF strategy or comparisons with SSD, YOLO9000, or YOLOv3. Searches for those terms found no passage in our copy that addresses the claimed performance ranking. (LLM, conf=0.76)
- Supportedcites [25] Redmon, 2018, YOLOv3: An incremental improvement source
The citation is attached to “YOLOv3” and identifies the framework used as a comparator. The paper’s Abstract (p. 1) presents YOLOv3 and reports its detection performance. It does not evaluate SA-CF, but that comparison is not specifically attributed to this citation. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
93 more claims · 9 with their cited sources checked
Create a free account to see all 227 claims and to verify the 48 references still open.
- It may state two separate claims as one. 59
- Probably not a substantial claim. 95
- The quote only partly states this claim. 1
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Systematic Review of Drone Based Road Traffic Monitoring System > C. PRE-PROCESSING APPROACHES
“the input images were segmented into a 4 × 4 block”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > C. PRE-PROCESSING APPROACHES
“adding the segments together with the original image, thus increasing the training set by 5× times”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Systematic Review of Drone Based Road Traffic Monitoring System > 1) For detection task
“recall refers to the same concept applied to all predictions, both positive and negative.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > 1) For detection task
“𝑅𝑒𝑐𝑎𝑙𝑙 = 𝑇 𝑃 𝑇 𝑃 + 𝐹 𝑁 (2)”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Systematic Review of Drone Based Road Traffic Monitoring System > 1) Traffic object detection
“multiple altitude values ranging from 5 -126 m have been considered when gathering data from various perspectives.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > VII. CONCLUSION
“| [53] Faster R-CNN | VGG-16 | Custom | 1920 × 1080 (24fps) | 100 to 150m |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ A Systematic Review of Drone Based Road Traffic Monitoring System > IV. AERIAL DATASETS AND EVALUATING METRICS (RQ1) A. DATASETS
“UAV Detection and Tracking (UAVDT) dataset [45] provides information for video object detection, SOT, and MOT tasks with vehicles as the target objects.”
§ A Systematic Review of Drone Based Road Traffic Monitoring System > 1) Traffic object detection
“VisDrone2021 [88] and DOTA [45] datasets”
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 3×
“Additionally, Sun et al. [110] investigated the performance of Faster R-733 CNN [83] and SSD [81] detectors with MobileNet [111], GoogleNet/Inception V2 [112] and ResNet-50 [88] backbone models for UAV videos.”
MethodFaster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Ren, S., He, K., Girshick, R., & Sun, J. (2017) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2016.2577031 · Cited by 73,472 · cited 1×
“with the advent of DL-based frameworks, including one-stage detectors [25], [26], as well as two-stage detectors [27], [28].”
MethodYou Only Look Once: Unified, Real-Time Object Detection
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.91 · Cited by 53,803
Going deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562 · cited 1×
“Additionally, Sun et al. [110] investigated the performance of Faster R-733 CNN [83] and SSD [81] detectors with MobileNet [111], GoogleNet/Inception V2 [112] and ResNet-50 [88] backbone models for UAV videos.”
MethodMicrosoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Dollár, P., & Zitnick, C. L. (2014) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-10602-1_48 · Cited by 45,460 · cited 1×
“In addition, in the majority of vehicle detection systems DL models are trained using weights of pre-trained models based on large scale datasets such as MS-COCO dataset [77], which contains regular and high resolution images in their natural context.”
MethodImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380
Histograms of Oriented Gradients for Human Detection
Dalal, N., & Triggs, B. (2005) · 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) · doi:10.1109/cvpr.2005.177 · Cited by 33,584
Feature Pyramid Networks for Object Detection
Lin, T.-Y., Dollar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.106 · Cited by 30,939 · cited 1×
“Furthermore, to address the real-time mon-705 itoring and occlusion problems in traffic monitoring sys-706 tem, [107] used FPN [108] and observed high precision at the cost of high computational complexity when considering high resolution images.”
MethodHe, K., Gkioxari, G., Dollar, P., & Girshick, R. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.322 · Cited by 29,137
Unpaired image-to-image translation using cycle-consistent adversarial networks
Zhu, J.-Y., Park, T., Isola, P., & Efros, A. A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.244 · Cited by 23,602 · cited 1×
“In this context, [75] proposed Faster R-CNN frame-589 work [83] with domain adaptation concept and using Gener-590 ative Adversarial Networks (GANs), specifically cycle GAN approach [90], annotated daytime images successfully turned into unlabelled nighttime images.”
Method
Research works printed but not linked: 7
- Masood, 2017, Trajectory tracking and hurdle avoidance for an autonomous quadrotor UAV, Eur. J. Adv. Eng. Technol., 4, 135
- Kaiming, 2017, Mask R-CNN, IEEE Trans. Pattern Anal. Mach. Intell., 42, 386
- Park, Large Scale Visual Recognition Challenge 2017
- 10.1109/TPAMI.2016.2577031 · DOI · Paper page
- Collins, An open source tracking testbed and evaluation web site, 2, 35
- Glenn, 2021, YOLOv5
- 2022, YOLOv6
Cited by
A Survey on UAV Applications in Smart City Management: Challenges, Advances, and Opportunities
Xu, H., Wang, L., Han, W., Yang, Y., Li, J., Lu, Y., & Li, J. (2023) · IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing · doi:10.1109/jstars.2023.3317500 · Cited by 118
“ment [15], [16], urban traffic management disaster [17], urban ecological environment monitoring [18], etc.”
BackgroundUAVs meet LLMs: Overviews and perspectives towards agentic low-altitude mobility
Tian, Y., Lin, F., Li, Y., Zhang, T., Zhang, Q., Fu, X., Huang, J., Dai, X., Wang, Y., Tian, C., et al. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103158 · Cited by 174
“In recent years, substantial research has been conducted in this area, including vehicle trajectory monitoring [408], road condition monitoring [409, 410], road-side units (RSUs) communication [411], and applications and management in urban scenarios [412].”
GPS Spoofing Attacks in FANETs: A Systematic Literature Review
Altaweel, A., Mukkath, H., & Kamel, I. (2023) · IEEE Access · doi:10.1109/access.2023.3281731 · Cited by 86
“…to the widespread acceptance of UAVs for various military and civilian applications such as border surveillance [2], disaster monitoring [3], traffic monitoring [4], managing wildfire [5], relay for Ad-Hoc networks [6], wind estimation [7], emergency management and high-risk situations [8], etc.”
BackgroundA Survey on Applications of Unmanned Aerial Vehicles Using Machine Learning
Teixeira, K. B., Miguel, G., Silva, H. S., & Madeiro, F. (2023) · IEEE Access · doi:10.1109/access.2023.3326101 · Cited by 79
“Bisio et al. [41] present a systematic review on drone-based traffic monitoring systems, particularly in the context of DL.”
BackgroundGC-YOLOv9: Innovative smart city traffic monitoring solution
An, R., Zhang, X., Sun, M., et al. (2024) · Alexandria Engineering Journal · doi:10.1016/j.aej.2024.07.004 · Cited by 61
Metadata
- Authors
- Bisio, I., Garibotto, C., Haleem, H., Lavagetto, F., & Sciarrone, A.
- Affiliations
- University of Genoa (IT)
- Venue
- IEEE Access
- Published
- 2022
- Volume
- 10
- Pages
- 101537-101555
- ISSN
- 2169-3536 (ISSN-L)
- OpenAlex
- W4296211548
- Cited by
- 139
- References
- 148
- Citations by year
- 2026: 41 · 2025: 49 · 2024: 27 · 2023: 21 · 2022: 1
- FWCI
- 1.76
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Locations
- IEEE Access (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- CINECA IRIS Institutial Research Information System (University of Genoa) (submittedVersion)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Video Surveillance and Tracking Methods
- Physical Sciences › Engineering › Automotive Engineering › Autonomous Vehicle Technology and Safety
- Keywords
- unmanned aerial vehicles, vehicle detection, vehicle tracking, vehicle counting, traffic congestion, traffic flow estimation, vehicle speed estimation, smart cities, deep learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A Systematic Review of Drone Based Road Traffic Monitoring System. https://app.cochat.ai/papers/10.1109/access.2022.3207282
- The paper
- Bisio, I., Garibotto, C., Haleem, H., Lavagetto, F., & Sciarrone, A. (2022). A Systematic Review of Drone Based Road Traffic Monitoring System. IEEE Access. https://doi.org/10.1109/access.2022.3207282
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.