AOW-YOLO: Ablation Study Results for Smoking Detection on Construction Sites

An ablation study for the AOW-YOLO model, a novel lightweight object detection system designed to identify smoking behavior in complex construction site environments. The dataset, shared by Ruishi Liang on figshare, contains experimental results showing the model's performance improvements over baseline models. It was last updated on May 6, 2026.

Use Cases

• Benchmarking lightweight object detection models based on reported mAP50 and mAP50:95 metrics.
• Analyzing the impact of novel loss functions like Adaptive Occlusion and Weighting IoU (AOWIoU) on model performance.
• Evaluating the efficiency of the Spatial Grouped-Pointwise (SGP) convolution module for feature integration in low-resolution, cluttered scenes.
• Researching methods to improve detection of small objects in complex backgrounds, as described for construction sites.

Strengths

• Model performance is quantitatively reported with specific metrics, including a 31.6% faster inference speed than YOLO11n.
• The dataset is shared under a permissive CC-BY-4.0 license, allowing for broad reuse and modification.
• The study is focused on a concrete, challenging application: detecting small smoking-related objects in cluttered construction site backgrounds.

Limitations

• The dataset is very small at 5.5 KB, suggesting it contains only summary results, not the underlying training or validation data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for statistical analysis.

Provenance

Source

figshare, author Ruishi Liang.

Collection Method

Likely contains tabular results from model ablation experiments.

Freshness

Last updated 2026-05-06 17:45:44; freshness should be verified.