Currently, we support comparing vector data predictions and groundtruths in utils.eval_confusion_matrix. Ideally, this should support raster data as either the prediction, ground truth, or both. Once the confusion matrix is built it may be helpful to pass it to the rastervision evaluation functionality to compute a comprehensive set of metrics.
Currently, we support comparing vector data predictions and groundtruths in
utils.eval_confusion_matrix. Ideally, this should support raster data as either the prediction, ground truth, or both. Once the confusion matrix is built it may be helpful to pass it to therastervisionevaluation functionality to compute a comprehensive set of metrics.