An Advanced Object Detection Framework for UAV Imagery Utilizing Transformer-Based Architecture and Split Attention Module: PvSAMNet
Bibliographic record
Abstract
In recent years, advancements in deep learning have fostered the development of sophisticated object detectors, specifically in the realm of computer vision.The inherent complexity of images captured by unmanned aerial vehicles (UAVs) presents a multitude of challenges for object detection.These include, but are not limited to, the detection of small and densely clustered objects, scale variance, occluded objects, and intricate backgrounds, which are particularly prevalent in drone-captured imagery when compared to natural scenes with larger and more distinct objects.The current landscape of object detection research has seen a surge in interest surrounding advanced, anchor-free object detectors, attention mechanisms, and the use of transformers as an alternative to convolutional neural networks.In light of these developments, this study introduces a novel object detection framework that eschews anchor utilization and leverages a transformer backbone for feature extraction.A cardinal grouping-based split attention module is integrated into this network to selectively extract the most pertinent features.The object detection head, termed the Pyramid Vision Split Attention Module Network (PvSAMNet), comprises three branches: classification, confidence, and regression, which collaboratively facilitate the final object detection from drone images.Additionally, an Intersection over Union (IoU) balanced loss function is employed to effectively equilibrate the classification and localization steps.The performance of the proposed detector is evaluated using the Visdrone-DET dataset, with the efficacy gauged by the average precision (AP) and average recall (AR) metrics.The results demonstrate that the proposed model outperforms other detector models with an average precision of 38.74.This study contributes to the ongoing discourse in the field of object detection, providing a novel framework that addresses the unique complexities of UAV imagery and demonstrates promising results in comparative evaluations.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".