Month: October 2019

Previous: #5 Features for Images in a Folder
Next: Not yet available

This is the sixth in a series of videos I am creating of the research I am performing under the direction of University of Nebraska Lincoln Professor Troy Gilmore to determine whether we can use images taken at the weir on North Platte River at the Nebraska-Wyoming State Line to replicate measurements of discharge and stage by extracting features from the images and using measurements from USGS sensors as ground truth.

This is the first pass of the weir finder. Notice that it starts missing at night and that it is pretty good, but not perfect. We will refine this.

Previous : #4 Features for Images in a Folder
Next: #6: Finding the Weir

This is the fifth in a series of videos I am creating of the research I am performing under the direction of University of Nebraska Lincoln Professor Troy Gilmore to determine whether we can use images taken at the weir on North Platte River at the Nebraska-Wyoming State Line to replicate measurements of discharge and stage by extracting features from the images and using measurements from USGS sensors as ground truth.

It is necessary to search images with odd (non-rectangular) regions of interest. Eventually, we want to be able to set those regions of interest automatically without manual intervention, but they will help greatly in algorithm development to get started.

Previous: #3 Add a Features Dialog Box
Next: #5 Odd shaped regions of interest

This is the fourth in a series of videos I am creating of the research I am performing under the direction of University of Nebraska Lincoln Professor Troy Gilmore to determine whether we can use images taken at the weir on North Platte River at the Nebraska-Wyoming State Line to replicate measurements of discharge and stage by extracting features from the images and using measurements from USGS sensors as ground truth.

Troy had me add the ability to calculate all the image features for a folder and show them in the GUI. This was a relatively minor change, but it works well. It should be noted that only rudimentary image features are being calculated at this point. We will add more, probably many more, features as we start building classifiers.

Previous: #2 Extracting Metadata from Images
Next: #4 Features for Images in a Folder

This is the third in a series of videos I am creating of the research I am performing under the direction of University of Nebraska Lincoln Professor Troy Gilmore to determine whether we can use images taken at the weir on North Platte River at the Nebraska-Wyoming State Line to replicate measurements of discharge and stage by extracting features from the images and using measurements from USGS sensors as ground truth.

Now that we can look at images, Troy wants a way to start adding features that are calculated from the images. This video is my first pass at adding the functionality we need. It is a dialog box that allows the user to see calculated features for a single image or from a folder of features. It presents the features for a folder of images in comma separated value (CSV) format. That will be used sometime in the future to feed the early machine learning.

Previous: #1 Infrastructure
Next: #3 Add a Features Dialog Box

This is the second in a series of videos I am creating of the research I am performing under the direction of University of Nebraska Lincoln Professor Troy Gilmore to determine whether we can use images taken at the weir on North Platte River at the Nebraska-Wyoming State Line to replicate measurements of discharge and stage by extracting features from the images and using measurements from USGS sensors as ground truth.

Troy’s second task for me was to extract the timestamps and other data from each of the image’s EXIF metadata. That task is now complete and is shown here as an addition to the WaterEval machine learning program for which we laid some groundwork shown in the previous video. We will use this timestamps to select the stage and discharge data from the USGS that applies to each of the images. In addition, we will see whether knowledge of the camera settings changes and can also be added as a feature for our machine learning model.