For more detail on Imagery sources see the previous Imagery section. There are multiple sources of satellite imagery out there (Maxar Secure Watch, Planet, list a few of them), and selecting your source of imagery depends on various factors including budget and timeframe. In creating your own training data, you have the most control over the quality of the data and can tailor the data exactly to your needs.Īcquiring Imagery is the first step in the training data process. With true optical flow.Your task is to acquire satellite imagery and create the building outline vector labels on the imagery, which will be used together to train the ramp model. The MPI Sintel Flow Dataset of synthetic video sequences annotated.Of satellite imagery co-sponsored by IARPA,Īnd Maxar (formerly DigitalGlobe and Radiant Solutions) The COIL-20 database and COIL-100 database of 20 and 100 objects respectively, imaged from 72 viewing directions, developed by Sameer Nene, Shree Nayar, and Hiroshi Murase.The Caltech 256 collections of images in 256 categories developed by Greg Griffin, Alex Holub, and Pietro Perona.The Caltech 101 collections of images in 101 categories developed by Fei-Fei Li, Rob Fergus, and Pietro Perona.The CIFAR-10 dataset of 60,000 images in ten classes developed by Alex Krizhevsky.NIST Special Database 19 of handprinted forms and characters (for purchase).The MNIST database of handwritten digits developed by Yann LeCun, Corinna Cortes, and Christopher J.PASCAL 3D+, a data set for 3D object detection in the wild developed by Yu Xiang, Roozbeh Mottaghi and Silvio Savarese.Made available by Stanford's ImageNet team/li> The ImageNet image download page (requires signup).Several tutorials on Jupyter notebooks can be found online.You can also find information by googling, but make sure you refer to version 3 of Python if you do so. Use the library reference and the language reference as your official sources of information about Python 3. The official Python 3 Documentation also includes a tutorial.Google's Python class is a leisurely but clear Python tutorial.The professional version has tools that are very useful for professional development but you won't need in this course. If you do a lot of programming outside this course, you may want to download the professional version, which is available for free here if you access that page from a Duke computer. You are urged to download the (free) P圜harm Integrated Development Environment (IDE). Any program that is longer than a few lines of code requires debugging, and debugging is a nightmare in a Python notebook.This distribution places all relevant files in the appropriate places, and you won't have to struggle with linking libraries, etc. This is essentially all you need for this course. This distribution includes Python, several basic libraries for data science (numpy, scipy, and more), visualization libraries (including matplotlib), machine learning libraries, including scikit-learn. It is strongly recommended that you use the latest version of Anaconda, rather than whatever you may have already on your computer. Anaconda is a stable and coherent distribution of Python for data science.
0 Comments
Leave a Reply. |