#DESIGN THE WEAVE FOR A SPACENET CODE#

The weights show the contribution of each successive layer of CosmiQNet v1. As the network grows (training deeper layers), the weights decrease - each deeper layer is a slight perturbation of the identity map.įigure 5: Layer Weights for CosmiQNet v1. The network is extended by optimizing a convex combination of the previous layer and the current layer, producing a weight for the new layer that measures its contribution to the final output.

The refinement is integrated into the training algorithms: where each additional layer is locally optimized, then all the previously layers are optimized together with the new layer. The philosophy of CosmiQNet is that each additional layer refines the output of the network.

If you want to hear more about these FCNNs, I plan to present at GTC DC 2016 for a different application. This architecture was developed for another application but seems to perform reasonably in this application. We include bypass parameter weights in each layer of CosmiQNet to facilitate training and to provide feedback on the necessary required depth. The resolution of the 8-band GeoTIFF is roughly one quarter (in each dimension) the resolution of the the 3-band GeoTIFF the difference in resolution is depicted by the scale of the GeoTIFFs. The inputs are at two resolutions and the output distance transform matches the lower of the input resolutions.

#DESIGN THE WEAVE FOR A SPACENET CODE#

The following unoptimized Python code performs the relabeling (for an implementation that uses the more efficient GDAL library see SpaceNetChallenge):įigure 1: Blackbox Fully Convolution Neural Network: CosmiQNet. Provided the building footprints do not overlap, this transform is invertible. This strategy transforms a label of building footprints into a raster image where each pixel has the value of the distance that pixel is to the boundary of a building - exterior pixels have a negative distance, interior pixels have a positive distance, the boundary of the footprint has a distance of zero. Since geoJSONs can be more descriptive than just bounding boxes, we employ a strategy that is more flexible with boundary descriptions than a rectangular box, see. Each polygon (or multi-polygon for non-simply connected regions) represents a building footprint.

The label contains a list of polygons contained within the sub-region stored in a geoJSON format.

The three-band image is derived from a panchromatic image and a subset of the three channels - the details of the derivation are available from the WorldView 2 specifications.

The 8-band raster image, at roughly 2 m ground sampling distance, contains both visible spectrum channels and near infrared channels with 16-bit values.

The 3-band raster image, at roughly 0.5 m ground sampling distance, contains Red, Green, and Blue color channels with 8-bit values.

For each sub-region, there are two images (GeoTIFFs) and one label (geoJSON): More information on SpaceNet is available here. The code snippets in this blog are shared under the following license:

We implement the metric derived from Intersection-over-Union to compute an F1-score for the machine learning algorithm.

We post-process the output to generate non-rectangular region proposals and output the proposed regions into geoJSON files.

We train a weighted fully convolutional neural network to label the output the weights give feedback on the importance of bands and layers.

We pre-process the labeled imagery using a distance transform to create a richer labeled data set.

The classifier uses 11 bands within the satellite imagery.

There are several properties of this challenge that distinguish it from other computer vision problems: Similar to many other computer vision problems, we desire to find objects (in particular, building footprints) in images (satellite imagery from SpaceNet). Sample Python code is included to demonstrate basic GIS functionality within Python and neural network design in TensorFlow.

The strategy employed is far from optimal but rather is meant to illustrate a complete workflow based on a previous post. This post presents a walk through of an object detection process applied to SpaceNet imagery.

Design the weave for a spacenet