Object Detection Prediction

Once the model is trained, you can run predictions on it or use it as a checkpoint to train further at a later date. This page requires a trained model (model.tflite or model.h5).

Simple Prediction API

The first thing you have to do import the necessary module:

from MAGIST.Vision.FullySupervisedModels.MAGIST_Lite_Detector import MAGIST_CNN, MAGIST_CNN_Predictor

Next, you must instantiate the class and provide the configuration file path.

predictor = MAGIST_CNN_Predictor("config.json")

Important

If you want the string IDs that correspond to the numerical answer the prediction provides, please look at the "Additional Methods" section

Single Image Prediction

To predict on a single image, you must use the appropriate method and provide the relative (or absolute) path to the image.

predictor.img_prediction("some_folder/another_folder/some_image.jpg")

More information coming soon