Learning Object-Class Segmentation with Convolutional Neural Networks
- Hannes Schulz ,
- Sven Behnke
Proceedings of European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning |
After successes at image classification, segmentation is the next step towards image understanding for neural networks. We propose a convolutional network architecture that includes innovative elements, such as multiple output maps, suitable loss functions, supervised pretraining, multiscale inputs, reused outputs, and pairwise class location filters. Experiments on three data sets show that our method performs on par with current computer vision methods with regards to accuracy and exceeds them in speed.