123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596 |
- # Config file for Stage 4 - Wave Detection
- # Name of stage, must be identical with folder name
- STAGE_NAME: 'stage04_wave_detection'
- # The profile name is the key for this parameter configuration. Results are stored in output_path/<PROFILE>/ (output_path is defined in settings.py)
- PROFILE: 'LENS|macrodim7'
- # Name of the output file
- STAGE_OUTPUT: "waves.nix"
- # File format in which all intermediate neo objects are stored
- NEO_FORMAT: 'nix'
- # If True (default), the output file of a stage is created as symbolic link
- # to the last block output. If False, a duplicate is created (e.g. for cloud
- # application, where sym-links are not supported).
- USE_LINK_AS_STAGE_OUTPUT: True
- # Plotting parameters
- PLOT_TSTART: 0 # in s
- PLOT_TSTOP: 10 # in s
- PLOT_CHANNELS: ['None', 'None', 'None'] # int or None. default 'None' -> randomly selected
- PLOT_FORMAT: 'png'
- # DETECTIION BLOCK
- ##################
- # Available Blocks: 'trigger_clustering'
- DETECTION_BLOCK: 'trigger_clustering'
- # ADDITIONAL PROPERTIES
- #######################
- # Available Blocks: 'optical_flow', 'critical_points', 'wave_mode_clustering'
- ADDITIONAL_PROPERTIES: ['wave_mode_clustering', 'optical_flow']
- # Wavefront Clustering
- ######################
- # Using sklearn.cluster.DBSCAN
- METRIC: 'euclidean'
- # eps, maximum distance between points to be neigbours
- NEIGHBOUR_DISTANCE: 3.21
- MIN_SAMPLES_PER_WAVE: 13
- # Factor from time dimension to space dimension in sampling_rate*spatial_scale
- TIME_SPACE_RATIO: 1.57 # i.e. distance between 2 frames corresponds to X pixel
- # Optical Flow (Horn-Schunck algorithm)
- ##############
- USE_PHASES: True
- # weight of the smoothness constraint over the brightness constancy constraint
- ALPHA: 1.5
- # maximum number of iterations optimizing the vector field
- MAX_NITER: 100
- # the optimization end either after MAX_NITER iteration or when the
- # maximal change between iterations is smaller than the CONVERGENCE_LIMIT
- CONVERGENCE_LIMIT: 0.0001
- # standard deviations for the Gaussian filter applied on the vector field
- # [t_std, x_std, y_std]. (0,0,0) for no filter
- GAUSSIAN_SIGMA: [0.5,0.79,0.79]
- # Kernel filter to use to calucualte the spatial derivatives.
- # simple_3x3, prewitt_3x3, scharr_3x3, sobel_3x3, sobel_5x5, sobel_7x7
- DERIVATIVE_KERNEL: 'scharr_3x3'
- # Critical Point Clustering
- ###########################
- # Wave Mode Clustering
- ######################
- # fraction of channels that need to be involved in a wave to be included
- MIN_TRIGGER_FRACTION: 0.5
- # number of similar waves to use to extrapolate missing trigger from
- NUM_WAVE_NEIGHBOURS: 5
- # percentage of similar waves to keep for the clustering
- WAVE_OUTLIER_QUANTILE: 1
- # number of pca dims to project the trigger patterns onto before clustering
- PCA_DIMS: 10
- # number of clusters for the kmeans algorithm
- NUM_KMEANS_CLUSTER: 6
- # grid spacing for the interpolation [0,1]
- INTERPOLATION_STEP_SIZE: 0.2857
- # smoothing factor (0: no smoothing)
- INTERPOLATION_SMOOTHING: 25
- # VIDEO SETTINGS
- ################
- QUALITY: 5 # 0(good) - 31(bad)
- SCALE_X: 720
- SCALE_Y: 720
- FPS: 10
- BITRATE: 20M
- # displayed sampling rate, the data will be stretched or compressed to.
- # If None, the inherent sampling rate is used.
- FRAME_RATE: None
- # 'gray', 'viridis' (sequential), 'coolwarm' (diverging), 'twilight' (cyclic)
- COLORMAP: 'twilight'
- PLOT_EVENT: 'wavefronts' # name of neo event to plot, default is None
- MARKER_COLOR: 'r'
|