Elamur Research
☰
home / papers / 2013 / Rich feature hierarchies for accurate object detection and semantic segmentation

Rich feature hierarchies for accurate object detection and semantic segmentation

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/1311.2524

Summary

Progress in object detection on PASCAL VOC had stalled around 2010–2012, with leading systems relying on hand-crafted features such as SIFT and HOG and on elaborate ensembles. Girshick et al. ask whether the CNN breakthrough in ImageNet classification carries over to detection. Their system, R-CNN, takes about 2000 class-agnostic candidate regions from selective search. It warps each region to a fixed 227×227 input and passes it through a Krizhevsky-style CNN (implemented in Caffe) to get a 4096-dimensional descriptor. Per-class linear SVMs then score each region.

To cope with the small amount of labelled detection data, the network is first trained for classification on ILSVRC2012 and then fine-tuned on PASCAL. A class-specific bounding-box regressor corrects localization errors. R-CNN reaches 53.3% mAP on VOC 2012, more than 30% better in relative terms than the previous best. It also beats OverFeat on ILSVRC2013 detection and gives competitive semantic segmentation results. The work established region-based deep detection and the pre-train-then-fine-tune recipe as standard practice.

Key points

  • VOC 2010: R-CNN reaches 53.7% mAP, compared with 33.4% for HOG-based DPM and 35.1% for the UVA system, which uses the same proposals with bag-of-visual-words features.
  • ILSVRC2013 detection (200 classes): R-CNN scores 31.4% mAP, against 24.3% for the sliding-window OverFeat detector.
  • Fine-tuning: Domain-specific fine-tuning after ILSVRC2012 pre-training adds about 8 mAP points.
  • Efficiency: Region features are shared across classes and are compact (4k dimensions, versus 360k for UVA). Proposals and features take about 13 s per image on a GPU, so the approach scales to thousands of classes.
  • Semantic segmentation: R-CNN reaches 47.9% average accuracy on the VOC 2011 test set. Later versions of the paper add results with the 16-layer Simonyan–Zisserman network.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Other

Cited by

238 papers of our library cite it (of the 140673 whose bibliographies we read); 25 of them have a page here, 213 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1311.2524 · arXiv 1311.2524

@misc{girshick2013rich,
  title = {{Rich feature hierarchies for accurate object detection and semantic segmentation}},
  author = {Girshick, Ross and Donahue, Jeff and Darrell, Trevor and Malik, Jitendra},
  year = {2013},
  doi = {10.48550/arxiv.1311.2524},
  eprint = {1311.2524},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1311.2524}
}

References

28 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. B. Alexe, T. Deselaers, and V. Ferrari. Measuring the objectness of image windows. TPAMI, 2012. 2 find
  2. P. Arbeláez, B. Hariharan, C. Gu, S. Gupta, L. Bourdev, and J. Malik. Semantic segmentation using regions and parts. In CVPR, 2012. 10, 11 find
  3. P. Arbeláez, J. Pont-Tuset, J. Barron, F. Marques, and J. Malik. Multiscale combinatorial grouping. In CVPR, 2014. 3 find
  4. J. Carreira, R. Caseiro, J. Batista, and C. Sminchisescu. Semantic segmentation with second-order pooling. In ECCV, 2012. 4, 10, 11, 13, 14 find
  5. J. Carreira and C. Sminchisescu. CPMC: Automatic object segmentation using constrained parametric min-cuts. TPAMI, 2012. 2, 3 find
  6. D. Cireşan, A. Giusti, L. Gambardella, and J. Schmidhuber. Mitosis detection in breast cancer histology images with deep neural networks. In MICCAI, 2013. 3 find
  7. N. Dalal and B. Triggs. Histograms of oriented gradients for human detection. In CVPR, 2005. 1 find
  8. T. Dean, M. A. Ruzon, M. Segal, J. Shlens, S. Vijayanarasimhan, and J. Yagnik. Fast, accurate detection of 100,000 object classes on a single machine. In CVPR, 2013. 3 find
  9. J. Deng, A. Berg, S. Satheesh, H. Su, A. Khosla, and L. Fei-Fei. ImageNet Large Scale Visual Recognition Competition 2012 (ILSVRC2012). http://www.image-net.org/challenges/LSVRC/2012/. 1 ↗
  10. J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. ImageNet: A large-scale hierarchical image database. In CVPR, 2009. 1 find
  11. J. Deng, O. Russakovsky, J. Krause, M. Bernstein, A. C. Berg, and L. Fei-Fei. Scalable multi-label annotation. In CHI, 2014. 8 find
  12. J. Donahue, Y. Jia, O. Vinyals, J. Hoffman, N. Zhang, E. Tzeng, and T. Darrell. DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition. In ICML, 2014. 2 find
  13. M. Douze, H. Jégou, H. Sandhawalia, L. Amsaleg, and C. Schmid. Evaluation of gist descriptors for web-scale image search. In Proc. of the ACM International Conference on Image and Video Retrieval, 2009. 13 find
  14. I. Endres and D. Hoiem. Category independent object proposals. In ECCV, 2010. 3 find
  15. M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman. The PASCAL Visual Object Classes (VOC) Challenge. IJCV, 2010. 1, 4 find
  16. C. Farabet, C. Couprie, L. Najman, and Y. LeCun. Learning hierarchical features for scene labeling. TPAMI, 2013. 10 find
  17. P. Felzenszwalb, R. Girshick, D. McAllester, and D. Ramanan. Object detection with discriminatively trained part based models. TPAMI, 2010. 2, 4, 7, 12 find
  18. S. Fidler, R. Mottaghi, A. Yuille, and R. Urtasun. Bottom-up segmentation for top-down detection. In CVPR, 2013. 4, 5 find
  19. K. Fukushima. Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological cybernetics, 36(4):193–202, 1980. 1 find
  20. R. Girshick, P. Felzenszwalb, and D. McAllester. Discriminatively trained deformable part models, release 5. http: //www.cs.berkeley.edu/˜rbg/latent-v5/. 2, 5, 6, 7 find
  21. C. Gu, J. J. Lim, P. Arbeláez, and J. Malik. Recognition using regions. In CVPR, 2009. 2 find
  22. B. Hariharan, P. Arbeláez, L. Bourdev, S. Maji, and J. Malik. Semantic contours from inverse detectors. In ICCV, 2011. 10 find
  23. D. Hoiem, Y. Chodpathumwan, and Q. Dai. Diagnosing error in object detectors. In ECCV. 2012. 2, 7, 8 find
  24. Y. Jia. Caffe: An open source convolutional architecture for fast feature embedding. http://caffe. berkeleyvision.org/, 2013. 3 find
  25. A. Krizhevsky, I. Sutskever, and G. Hinton. ImageNet classification with deep convolutional neural networks. In NIPS, 2012. 1, 3, 4, 7 find
  26. Y. LeCun, B. Boser, J. Denker, D. Henderson, R. Howard, W. Hubbard, and L. Jackel. Backpropagation applied to handwritten zip code recognition. Neural Comp., 1989. 1 find
  27. Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition. Proc. of the IEEE, 1998. 1 find
  28. J. J. Lim, C. L. Zitnick, and P. Dollár. Sketch tokens: A learned mid-level representation for contour and object detection. In CVPR, 2013. 6, 7 find