Options JSON

Options JSON files can be saved with help of Aydin Studio(GUI). They can be passed to Aydin CLI to run on many more images with same options easily.

We basically have a python dict to store chosen options in the background. When save options JSON functionality is triggered, we basically encode the dict with help of jsonpickle package and we dump it into a new .json file. To understand the format of options JSON, you can inspect the example shared below:

"{
    \"feature_generator\": {
        \"class\": {
            \"py/type\": \"aydin.features.standard_features.StandardFeatureGenerator\"
        },
        \"kwargs\": {
            \"dct_max_freq\": 0.5,
            \"decimate_large_scale_features\": true,
            \"dtype\": {
                \"py/reduce\": [
                    {\"py/type\": \"numpy.dtype\"},
                    {\"py/tuple\": [\"f4\", false, true]},
                    {\"py/tuple\": [3, \"<\", null, null, null, -1, -1, 0]}
                ]
            },
            \"extend_large_scale_features\": false,
            \"include_corner_features\": false,
            \"include_dct_features\": false,
            \"include_fine_features\": true,
            \"include_line_features\": false,
            \"include_median_features\": false,
            \"include_random_conv_features\": false,
            \"include_scale_one\": false,
            \"include_spatial_features\": false,
            \"max_level\": 13,
            \"min_level\": 0,
            \"num_sinusoidal_features\": 0,
            \"scale_one_width\": 3,
            \"spatial_features_coarsening\": 2
        }
    },
    \"it\": {
        \"class\": {
            \"py/type\": \"aydin.it.fgr.ImageTranslatorFGR\"
        },
        \"kwargs\": {
            \"balance_training_data\": false,
            \"favour_bright_pixels\": false,
            \"max_voxels_for_training\": null,
            \"voxel_keep_ratio\": 1.0
        }
    },
    \"regressor\": {
        \"class\": {
            \"py/type\": \"aydin.regression.cb.CBRegressor\"
        },
        \"kwargs\": {
            \"compute_load\": 0.95,
            \"gpu\": true,
            \"gpu_devices\": null,
            \"learning_rate\": 0.01,
            \"loss\": \"l1\",
            \"max_bin\": null,
            \"max_num_estimators\": 2048,
            \"min_num_estimators\": 512,
            \"num_leaves\": 512,
            \"patience\": 32
        }
    },
    \"variant\": \"Noise2SelfFGR-cb\"
}"