4.2.2 Using Orthomosaic Maps
PicStork supports the COG Compatible Orthomosaic map files for creating a training task.
COG: An online storage and delivery format for georeferenced raster imagery is called a cloud-optimized GeoTIFF (COG). It's a typical GeoTIFF file that's been prepared for use in a cloud-based setting, like a web server or cloud storage service.
COGs enable Raster Data to be steamed across the internet without the need for a specialized GIS Server.
Lossless compression is supported by the popular raster image format TIFF (Tagged Image File Format).
GeoTIFF is a type of TIFF that contains geographic metadata, enabling the image to be precisely positioned on the globe.
A version of GeoTIFF designed specifically for use in a cloud environment is called COG (Cloud Optimised GeoTIFF). However, it can be used in the same ways as a regular GeoTIFF!
Training tasks can be executed by using COG-compatible orthomosaic maps also. This section will discuss the stepwise execution of the same.
Create a task: To create any type of task, firstly the user has to create a task with the specific name and project.
Create a Project: For handling various tasks created, the user has to create a project. Either user can select an existing project or create a new project depending on the requirements of an application.
Create an Imageset: To create an imageset click on the desired format of images from the options available on the screen.
For creating training tasks, the user can upload jpg, png files or orthomosaic COG tiff files can be uploaded. The user can import the map file from DroneNaksha directly. Select the appropriate option from the given options.
The map will get opened on the screen with the orthomosaic map generated through DroneNaksha as shown in image below.
Annotate the Images: Allows the images by adding the classes to annotate with the following tools. PicStork provides a powerful tool for the annotation of maps as multiple shape annotation. This functionality is available for annotating the maps in training tasks.
Select: Select the annotated object to edit, copy, and delete.
Box: Annotate the object with a box shape.
Polygon: Annotate the image with a polygon shape. The annotation of non-symmetric, non-regular objects is made easier by polygons. An object's dimensions are marked with dots, and the perimeter or circumference is manually drawn with lines.
Circle: Annotate the image with a circle shape. Circle annotation can be very useful when an object is perfectly circular. It will save a lot of time compared to manually drawing or using a polygon to create a circle.
Re-center: To re-center the map on the screen this option is used. This functionality will be useful when the user zooms in many times and the map or image is not completely visible on the screen.
The user can export this data by using the "Export" button. All the annotations can be cleared by clicking on the "Clear" button. After finishing the process of annotation, the user may click on the "Done" button for the next process.
PicStork gives details of annotation based on classes used. The user can verify it and perform some editing if required.
Edit the Annotations:
The user can edit the selected annotations. For editing the annotations, firstly select the object by the "Select" tool and then click on the object, now the user will have the options to edit the annotation such as:
Copy: Select the object to copy and the user marks the copies of such similar objects.
Delete: Select the object and delete it if you want to remove it.
Edit: Edit the annotation using this option. Select the object and select the edit option that appears on the screen. Editing of object facility allows the user to change the shape, and alter the position of the object.
Concept of Editing of annotations will be clear through this screen recording.
Setup:
Click on the "Setup" button to setup for training task. Ensure that all the images are fully annotated.
Select Processing Parameters for accuracy level and select the purpose for training.
PicStork provides 3 options for selecting the accuracy level for training Low with faster processing, Medium with slower, and High with slowest processing time. Select the accuracy level according to the requirement of the application.
Select the purpose of training as "Counting" and "Defect".
Counting: This type of task is used to count something like counting the palm trees from a farm.
Defect: This type of task is used to find out the defects if any.
The report generated from these two tasks will be separate. (refer to section 6.3 of the same user manual for a detailed discussion of reports).
Annotating Classes:
Add an orthomosaic image for annotation for creating a training task.
Annotate the image with new classes added as shown in the image.
The annotations marked in the training and testing area will be considered for training to create a machine learning model. The detector created from this process will run on the accuracy area to find out the accuracy of the objects based on algorithm and detector created. The annotations made in the accuracy area will be considered for calculating the accuracy of the detectors applied to find out the objects. If the same objects are available then the accuracy is better whereas if the objects are with different sizes then the accuracy of these objects will get affected.
After proper annotations, proceed for selecting processing parameters for accuracy level of training and training purpose.
Task details will be displayed as shown in the above image. The training process will take some time to complete. After completing a task, the accuracy of detector will be displayed in percentage with other details.
Last updated
