Surface Normal Estimator With Synthetic Data

This dataset supports the 'Three-Filters-to-Normal' method for accurate and fast surface normal estimation, as presented at RAL+ICRA 2021. It is associated with synthetic data, disparity maps, and depth information for training and evaluation. The author is ruirangerfan, and the project was last updated in January 2022.

Use Cases

• Train a surface normal estimation model using synthetic data and disparity maps.
• Evaluate the accuracy of depth-to-normal conversion algorithms on provided depth data.
• Benchmark ultrafast surface normal estimators against the method described in the RAL+ICRA 2021 paper.

Strengths

• The associated research was published at the RAL+ICRA 2021 conference.
• The project source code is distributed under a permissive MIT license.
• The repository was last updated on January 12, 2022.

Limitations

• The specific dataset size, row count, column structure, and file formats are unknown.
• The data appears to be primarily synthetic, which may limit generalizability to real-world scenes.
• No sample data is available for immediate inspection of data quality and format.

Provenance

Source

GitHub repository by author ruirangerfan.

Collection Method

Likely generated synthetically for the research on surface normal estimation.

Freshness

Last updated 2022-01-12.