2Ddesign-to-label-1.8.3 - Binary Map Antenna Designs with Operating Frequency Range Labels

This dataset is associated with the publication titled "A Convolutional Neural Network for Predicting the Frequency Range of Antennas Based on Randomized Designs" by Bartosz Czaplewski. The article is currently under review. The dataset will be publicly available upon the publication of the article. The dataset contains: I. Pseudo-random antenna designs as binary maps (as a numpy array in data_v1_generator_v8_dim_103_scale_all_DESIGNS_v3_maps_256.npy). II. Metadata corresponding to the above antenna designs, including (as columns in data_v1_generator_v8_dim_103_scale_all_METADATA_v3.csv): bandwidth_ranges - Frequency ranges where the antenna’s reflection response is below 3dB, num_bandwidths - The number of above-mentioned detected frequency ranges, center_frequencies - The center frequencies of the above-mentioned frequency ranges, bandwidth_spans - Absolute bandwidths of the above-mentioned frequency ranges, relative_bandwidth_spans - Relative bandwidths of the above-mentioned frequency ranges in relation to the center frequencies, bandwidth_classes - Assigned labels representing operating frequency ranges. Each label corresponds to one of the following frequency bands: 3.0–3.5 GHz, 3.5–4.0 GHz, ..., 7.5–8.0 GHz, num_classes - The number of assigned labels, 3.0-3.5GHz - Binary representation of the assignment of the "3.0-3.5GHz" label (0 = no label, 1 = label assigned), 3.5-4.0GHz - Binary representation of the assignment of the "3.5-4.0GHz" label (0 = no label, 1 = label assigned), 4.0-4.5GHz - Binary representation of the assignment of the "4.0-4.5GHz" label (0 = no label, 1 = label assigned), 4.5-5.0GHz - Binary representation of the assignment of the "4.5-5.0GHz" label (0 = no label, 1 = label assigned), 5.0-5.5GHz - Binary representation of the assignment of the "5.0-5.5GHz" label (0 = no label, 1 = label assigned), 5.5-6.0GHz - Binary representation of the assignment of the "5.5-6.0GHz" label (0 = no label, 1 = label assigned), 6.0-6.5GHz - Binary representation of the assignment of the "6.0-6.5GHz" label (0 = no label, 1 = label assigned), 6.5-7.0GHz - Binary representation of the assignment of the "6.5-7.0GHz" label (0 = no label, 1 = label assigned), 7.0-7.5GHz - Binary representation of the assignment of the "7.0-7.5GHz" label (0 = no label, 1 = label assigned), 7.5-8.0GHz - Binary representation of the assignment of the "7.5-8.0GHz" label (0 = no label, 1 = label assigned). III. An example of loading and displaying antenna designs written in Python (read_binary_map_designs_from_file.py). IV. An example of loading and displaying label vectors written in Python (read_labels_from_file.py). Details regarding antenna designs as binary maps and the labeling process are described in the publication titled "A Convolutional Neural Network for Predicting the Frequency Range of Antennas Based on Randomized Designs" by Bartosz Czaplewski. The link and full bibliographic description will be added after the publication of the paper. Moreover, files data_v1_generator_v8_dim_103_scale_all_DESIGNS_v3_maps_256_classless_30k.npy and data_v1_generator_v8_dim_103_scale_all_METADATA_v3_classless_30k.csv contain analogous datasets, but for samples without assigned labels. Extra files: Source code and pretrained models:CNN_2Ddesign-to-label_training_v6.py - script for training the modelmodels.py - file containing model definitionstoolkit.py - file containing additional functionsmodel_CNN_2Ddesign-to-label_20250405193431.pth - pretrained model for dataset with extra 30k classless samplesmodel_CNN_2Ddesign-to-label_20250210133755.pth - pretrained model for dataset with no classless samples