Research Article
Recursive Neural Networks Based on PSO for Image Parsing
Table 1
Accuracy of pixel accuracy of state-of-the-art methods on Stanford background dataset.
| Method and semantic pixel accuracy in % |
| Pixel CRF, Gould et al. (2009) 74.3 | Log. Regr. on superpixel features 75.9 | Region-based energy, Gould et al. (2009) 76.4 | Local labeling, Tighe and Lazebnik (2010) 76.9 | Superpixel MRF, Tighe and Lazebnik (2010) 77.5 | Simultaneous MRF, Tighe and Lazebnik (2010) 77.5 | Traditional RNN, Socher and Fei-Fei (2011) 78.1 | PSO-based RNN (our method) 78.3 |
|
|