code

Superpixel Segmentation using Depth Information / Code

The code provided as part of the bachelor thesis "Superpixel Segmentation using Depth Information" [1]:

• Revised implementation of the SEEDS superpixel algorithm [2]: SEEDS Revised.
• Superpixel library containing several of the evaluated algorithms: Superpixels Revisited.
• Benchmark used for evaluation of all superpixel algorithms: Extended Berkeley Segmentation Benchmark.
• Tool for working with the NYU Depth Dataset V2 [3]: NYU Depth V2 Tools.
• Interactive plots of comparison based on nvd3: nvd3 Superpixel Comparison.

Usage

Evaluation on the Berkeley Segmentation Dataset (BSDS500) [4] or on the NYU Depth Dataset V2 (NYUV2) [3]:

1. Download the BSDS500 or the NYUV2.
   • The NYUV2 dataset was split into training and test set according to nyu-depth-v2-tools/list_train.txt and nyu-depth-v2-tools/list_test.txt, respectively. Then subsets were selected according to nyu-depth-v2-tools/list_train_subset.txt and nyu-depth-v2-tools/list_test_subset.txt (using nyu-depth-v2-tools/convert_dataset.m, nyu-depth-v2-tools/collect_train_subset.m and nyu-depth-v2-tools/collect_test_subset.m). All images were cropped using nyu-depth-v2-tools/crop_dataset.m.
   • The BSDS500 already provides training and test set.
2. Compile all superpixel algorithms using the instructions in superpixels-revisited/README.md.
3. Set up the extended version of the Berkeley Segmentation Benchmark [5] according to extended-berkeley-segmentation-benchmark/README.md.

All command line tools in superpixels-revisited provide a --csv option in order to save all generated superpixel segmentations in the form of CSV files (also use --output my_output_folder/). The extended Berkeley Segmentation Benchmark provides extended-berkeley-segmentation-benchmark/convert_csv_bsd.m used to convert CSV files to the BSDS500 groundtruth format. For further instructions on using the extended Berkeley Segmentation Benchmark, see extended-berkeley-segmentation-benchmark/README.md.

File Index

• SEEDS Revised
• Superpixels Revisited
• Extended Berkeley Segmentation Benchmark
• NYU Depth V2 Tools
• nvd3 Superpixel Comparison

References

[1] D. Stutz, A. Hermans, B. Leibe.
    Superpixel Segmentation using Depth Information.
    Bachelor thesis, RWTH Aachen University, Aachen, Germany, 2014.
[2] M. van den Bergh, X. Boix, G. Roig, B. de Capitani, L. van Gool.
    SEEDS: Superpixels extracted via energy-driven sampling.
    European Conference on Computer Vision, pages 13–26, 2012.
[3] N. Silberman, D. Hoiem, P. Kohli, R. Fergus.
    Indoor segmentation and support inference from RGBD images.
    In Computer Vision, European Conference on, volume 7576 of Lecture Notes in Computer Science, pages 746–760. Springer Berlin Heidelberg, 2012.
[4] P. Arbeláez, M. Maire, C. Fowlkes, J. Malik. 
    Contour detection and hierarchical image segmentation.
    Transactions on Pattern Analysis and Machine Intelligence, 33(5):898–916, 2011
[5] P. Arbeláez, M. Maire, C. Fowlkes, J. Malik.
    Contour detection and hierarchical image segmentation.
    Transactions on Pattern Analysis and Machine Intelligence, volume 33, number 5, pages 898–916, 2011.

License

See the corresponding subrepositories for license information.