MSCNet Module Impact Analysis for LiDAR Segmentation

Experimental results evaluate a multi-scale LiDAR semantic segmentation network, MSCNet, across three public datasets. The analysis, authored by Xuewen Feng, was last updated in April 2026. The 5.5 KB file contains performance metrics comparing the model's parameters, accuracy, and running time against other methods.

Use Cases

• Compare segmentation accuracy metrics across the SemanticKITTI, SemanticPOSS, and Pandaset datasets to benchmark model performance.
• Analyze the trade-off between model parameter count and running time to identify architectures suitable for real-time applications.
• Evaluate the impact of specific modules like the multi-scale feature fusion block or pyramid pooling module on overall segmentation results.

Strengths

• Results are benchmarked on three established public datasets: SemanticKITTI, SemanticPOSS, and Pandaset.
• Explicitly reports metrics for parameter count, accuracy, and running time, enabling direct comparison.

Limitations

• The dataset is extremely small at 5.5 KB, containing only summary results, not raw LiDAR data or model weights.
• No column names or sample data are provided, limiting detailed reproducibility of the analysis.

Provenance

Source

Xuewen Feng, published on figshare.

Collection Method

Experimental results from training and evaluating the proposed MSCNet neural network.

Freshness

Last updated in April 2026.