Skip to content

classes

Jump to bottom
Anne Draelos edited this page Oct 31, 2018 · 3 revisions

OnACID

Attributes

  • params: CNMFParams

  • estimates: Estimates

  • N: int from columns in estimates.A, # of imaged sources

  • M: int from params and self.N, # of total components (background + imaged)

  • dims: (int,int) dimensions of FOV in pixels

  • ind_A: index matrix of Ab stacked matrix; separate from estimates

  • update_counter: array counter for updating shapes, for each component

  • time_neuron_added: array of 'when' component are added

  • time_spend: cumulative time spent running

  • loaded_model: keras model from json; optional

  • Ab_dense_copy: from estimates.Ab_dense. Unused.

  • Ab_copy: from local variable Ab_ (copied from estimates.Ab), then copied back into estimates.Ab after shape updating

  • Ab_epoch: copies of estimates.Ab for each epoch

  • bnd_Y: percentiles of loaded data; used if show movie.

  • bnd_AC: same for A.C

  • bnd_BG: same for b.f. Unused.

  • img_min: min of loaded data

  • img_norm: norm of loaded data, std+median

  • t: counter during online fitting; used if show movie and in create_frame method (problematic)

  • t_shapes: list of times for shape updating

  • t_detect: list of times

  • t_motion: list of times

  • comp_upd: list of updated components, used to count updating shapes during fit_next

  • captions: list of captions in show movie. Unused.

  • dview: Unused.

Below from update_num_components: Potentially problematic.

  • rhos_buf: RingBuffer, after update, from estimates.rho_buf. Unused.
  • ind_new_all: Unused.
  • cnn_pos: Unused.

Methods

  • fit_next: fits the next frame, updates the object
  • initialize_online: initialize using small portion of the dataset
  • fit_online: take files and fit in real-time
  • create_frame: currently only used for showing movie, has implicit assumptions
  • _prepare_object: prepares the online object given some estimates
  • save: save object in hdf5 (h5) format

Other methods in online_cnmf:

  • bare_initialization: bypass cnmf to quickly init OnACID (default)
  • seeded_initialization: init OnACID from user specified binary masks
  • HALS4shape: reshape A
  • HALS4activity: get C using block-coordinate descent
  • demix_and_deconvolve: get C using OASIS within b-c descent
  • init_shapes_and_sufficient_stats: estimate shapes on initial batch
  • update_shapes: updates shapes
  • update_num_components: check for new components in residual buffer, adds if needed
  • get_candidate_components: extract new candidate components and test them
  • remove_components_online: remove indexed components
  • initialize_movie_online: init movie using cnmf
  • load_OnlineCNMF: load object from save (hdf5)
  • csc_append: appends second csc_matrix to the right of the first one
  • corr: fast correlation
  • rank1nmf: fast rank 1 NMF

RingBuffer

Implements ring buffer efficiently, inherits from np.ndarray

Attributes

  • max_, cur

Methods

  • append, get_ordered, get_first, get_last_frames

CNMFParams

Class for setting processing parameters, grouped. Dictionary implementation

Attributes

  • Primary dicts: data, patch, preprocess, init, spatial, temporal, merging, quality, online, motion Should all have default values for each specified key.

Methods

  • set: add key-value pairs to a given dict (group)
  • get: gets value from group.key
  • get_group: gets full dict of given group
  • change_params: given dict, set all new values
  • to_dict: convert all dicts to single large dict
  • eq: define comparison for dicts (dict_compare method defined in .utilities)

Clone this wiki locally