Building Models

Could you elaborate on the process of pre-training models within PicStork and the procedure for adding objects for identification?


Hello @Anand.Nayak ,
Picstrok essentially offers model training and executes detection using the generated model.

Here are the steps for training the model:

  1. Begin by creating a project and task for model training.
  2. Subsequently, establish an image set and upload images. If you have an orthomosaic map created on the DroneNakhsa, you can also import it into the image set.
  3. Initiate the task by annotating the required objects. If the training task involves images, annotate a minimum of 20 images. Alternatively, if you’re working with an orthomosaic map, draw the Training area, Testing area, and Accuracy area.
  4. Proceed to the setup option, where you will select task accuracy, and task purpose, and commence the training task.
  5. Once the model is ready, you will receive an email notification.
  6. Verify your created model in the detector section of the Picstrok dashboard.stance.
1 Like