- Image processing is a method to perform some operations on an image, in order to get an enhanced image or to extract some useful information from it. It is a type of signal processing in which input is an image and output may be image or characteristics / features associated with that image.
-
Image Segmentation is a pixel level classification of an image. Road area or Building feature extraction can be considered as a segmentation or pixel-level classification problem.
-
Neural Networks have already been used in various remote sensing tasks, like damage assessment of washed-away building from pre and post tsunami aerial images. Furthermore, researchers are using Deep Learning to solve the problems of interpretation and understanding of remote sensing data.
-
Road Segmentation from Aerial Imagery is a challenging task. Obstruction from nearby trees, shadows of adjacent buildings, varying texture and color of rooftops, varying shapes and dimensions of buildings are among other challenges that hinder present day models in segmenting sharp building boundaries.
-
High-quality aerial imagery datasets facilitate comparisons of existing methods and lead to increased interest in aerial imagery applications in the machine learning and computer vision communities.
-
The Massachusetts Roads Dataset consists of 1171 aerial images of the state of Massachusetts. Each image is 1500 × 1500 pixels in size, covering an area of 2.25 square kilometers.
-
We randomly split the data into a training set of 1108 images, a validation set of 14 images and a test set of 49 images. The dataset covers a wide variety of urban, suburban and rural regions and covers an area of over 2600 square kilometers.
-
The test set alone covers over 110 square kilometers. The target maps were generated by rasterizing road centerlines obtained from the OpenStreetMap project.
-
A line thickness of 7 pixels and no smoothing was used in generating the labels. All imagery is rescaled to a resolution of 1 pixel per square meter.
- This dataset is derived from Volodymyr Mnih's original Massachusetts Buildings Dataset. Massachusetts Roads Dataset & Massachusetts Buildings dataset were introduced in Chapter 6 of his PhD thesis.
-
Given enough training time on GPU, larger datasets can be used for training the network. This can highly improve the accuracy and allow leveraging such image / semantic segmentation techniques in production Enterprise GIS environments. Rapid advances in Image Understanding using Computer Vision techniques have brought us many state-of-the-art deep learning models across various benchmark datasets.
-
Deep learning is a machine learning technique that teaches computers to do what comes naturally to humans: learn by example. It is getting lots of attention lately and for good reason. It’s achieving results that were not possible before. These models can achieve state-of-the-art accuracy, sometimes exceeding human-level performance.
Thank you!