How to re-train

i need to understand the full mechanism of retraining the the model
is retraining images with false positive detection?

To retrain a model, all you need to do (2 options depending on where you train it):

  1. Retrain a custom dataset on a model of your choice from our Model Library: copy the export code (same workspace, same project, and same version) and paste it in the desired notebook (a copy of the notebook you used the first time you trained a model on the dataset)

  2. Retrain a Roboflow model:

  • Generate a version of your dataset with the exact same preprocessing and augmentation settings as the version you want to “retrain.” (the images and annotations within the dataset must also be unchanged from the generated dataset version you would like to compare your results to)
  • Select a model type for training: Choosing Between Computer Vision Model Sizes
  • Train from Scratch if you did that the first time, or Train from Checkpoint (and from the exact same checkpoint) if you did that the first time (note: “the first time” is in reference to the dataset version you would like to compare results with)