A fast learning algorithm for image segmentation with max-pooling convolutional networks

By Jonathan Masci, Alessandro Giusti, Dan Claudiu Ciresan, Gabriel Fricout, J. Urgen Schmidhuber. 2013

78 citations recorded by the source corpus.

We present a fast algorithm for training MaxPooling Convolutional Networks to segment images. This type of network yields record-breaking performance in a variety of tasks, but is normally trained on a computationally expensive patch-by-patch basis. Our new method processes each training image in a single pass, which is vastly more efficient. We validate the approach in different scenarios and report a 1500-fold speed-up. In an application to automated steel defect detection and segmentation, we obtain excellent performance with short training times.

Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Industrial and Manufacturing Engineering

Citations in the corpus

From image-level to pixel-level labeling with Convolutional Networks (2014, methodology): “As shown in [15, 19], one can efficiently retrieve the label of all pixels of the image using a CNN model, by simply shifting the input image in both spatial directions, and forwarding it again through the network.”
Deep MRI brain extraction: A 3D convolutional neural network for skull stripping (2016, unknown): “For predictions, we feed the CNN with volumes of size 1803 voxels, resulting in 1283 predicted voxels in one forward pass utilizing a technique called maxfragment pooling (Masci et al., 2013).”
Weakly Supervised Semantic Segmentation with Convolutional Networks (2014, methodology): “As shown in [13, 17], one can efﬁciently retrieve the label of all pixels of the image using a CNN model, by simply shifting the input image in both spatial directions, and forwarding it again through the network.”
My First Deep Learning System of 1991 + Deep Learning Timeline 1962-2013 (2013, methodology): “ArcelorMittal, the world’s top steel producer, is using our methods for material defect detection and classiﬁcation, e.g., [63].”
Image Super-Resolution with Fast Approximate Convolutional Sparse Coding (2014, background): “This property is often overlooked (see [14, 15] for some notable recent exceptions), yet is one of the main ingredients in order to allow the transfer of learned patch-based super-resolution to full image superresolution.”
Change Detection in Synthetic Aperture Radar Images Based on Deep Neural Networks (2016, background): “With the development of deep learning, both the frameworks and training algorithms of CNNs have been improved [22], [23].”