8a600f790ba68880e9a76a5eda33a3f4fccd22c4,luminoth/utils/predicting.py,,get_predictions,#Any#Any#,32

Before Change

            "objects_labels": preds["objects_labels"],
            "objects_labels_prob": preds["objects_labels_prob"],
            "inference_time": preds["inference_time"],
            "scale_factor": preds["scale_factor"],
            "image_path": image_path,
        })

After Change

    We load the checkpoint in the first iteration and then use the same
    session and graph for subsequent images.
    
    config = get_config(config_files)

    // Don"t use data augmentation in predictions
    config.dataset.data_augmentation = None

Instances