Ablation Experiment Results on UAVid Semantic Segmentation

A collection of results from ablation experiments on the UAVid dataset for a semantic segmentation network called CRDFNet. The experiments validated the network's performance on remote sensing imagery, which is characterized by complex boundary shapes and dense small targets. The data is stored in an XLS file with a size of 5.5 KB.

Use Cases

• Compare F1 score, OA, and mIoU metrics from ablation experiments to evaluate the impact of different network modules.
• Analyze segmentation performance on dense small targets using results from the dual attention feature refinement module (DAFRM).
• Assess the effectiveness of the channel feature aggregation module (CFAM) in handling complex boundary shapes in remote sensing images.

Strengths

• Results are validated on established remote sensing datasets including UAVid, Potsdam, Vaihingen, and MSIDBG.
• The dataset is shared under a permissive CC BY 4.0 license.
• The file is small (5.5 KB), facilitating quick access and analysis.

Limitations

• The dataset is very small (5.5 KB), indicating limited scope and likely a summary of results rather than raw experimental data.
• Specific metrics like row count, column names, and sample data are unavailable, limiting detailed reproducibility.
• The data format is XLS, which may not be optimal for large-scale programmatic analysis common in machine learning.

Provenance

Source

Author Xin Wang, published on figshare.

Collection Method

Generated from ablation experiments on a proposed semantic segmentation network (CRDFNet).

Freshness

Last updated on 2026-03-20.