5/28/2023 0 Comments Convert image format in potLabels_file - path to file with word description of labels (synset_words). Imagenet - convert ImageNet dataset for image classification task to ClassificationAnnotation.Īnnotation_file - path to annotation in txt format. Optional, more details in Customizing dataset meta section.įashion_mnist - convert Fashion-MNIST dataset to ClassificationAnnotation.Īnnotation_file - path to labels file in a binary format.ĭata_file - path to images file in a binary format. Mnist_csv - convert MNIST dataset for handwritten digit recognition stored in csv format to ClassificationAnnotation.Īnnotation_file - path to dataset file in csv format.Ĭonvert_images - allows converting images from annotation file to user specified directory (default value is False).ĭataset_meta_file - path to json file with a dataset meta (e.g. Optional, more details in Customizing dataset meta section. Images_file - binary file which contains images.Ĭonvert_images - allows converting images from data file to user specified directory (default value is False).Ĭonverted_images_dir - path to converted images location if enabled convert_images.ĭataset_meta_file - path to json file with dataset meta (e.g. Labels_file - binary file which contains labels. Mnist - convert MNIST dataset for handwritten digit recognition to ClaassificationAnnotation. label_map, color_encoding).Optional, more details in Customizing dataset meta section. Num_classes - the number of classes in the dataset - 10 or 100 (Optional, default 10)ĭataset_meta_file - path to json file with dataset meta (e.g. Has_background - allows to add background label to original labels (Optional, default value is False).Ĭonvert_images - allows converting images from pickle file to user specified directory (default value is False).Ĭonverted_images_dir - path to converted images location. test_batch).īatch_meta_file - path to pickle file which contains label names (e.g. Default value is False.Īccuracy Checker supports following list of annotation converters and specific for them parameters:Ĭifar - converts CIFAR classification dataset to ClassificationAnnotationĭata_batch_file - path to pickle file which contain dataset batch (e.g. Supported annotations: ClassificationAnnotation, DetectionAnnotation, MultiLabelRecognitionAnnotation, RegressionAnnotation. You can use this parameter if you need to reuse converted annotation to avoid subsequent conversions.ĭataset_meta - path to store meta information about converted annotation if it is provided.Īnalyze_dataset - flag which allow getting statistics about the converted dataset. You can also specify subsample_seed if you want to generate subsample with specific random seed.Īnnotation - path to store converted annotation pickle file. Please, be careful to use this option, some datasets does not support subsampling. You can specify the number of ground truth objects or dataset ratio in percentage. You can additionally use optional parameters like: All paths can be prefixed via command line with -s, -source argument. Each conversion configuration should contain converter field filled with a selected converter name and provide converter specific parameters (more details in supported converters section). Describing Annotation Conversion in Configuration File ¶Īnnotation conversion can be provided in dataset section of your configuration file to convert annotation in-place before every evaluation.
0 Comments
Leave a Reply. |