Skip to content

Process Raw Frames

Jump to bottom
chaichontat edited this page Aug 15, 2019 · 3 revisions

Pipeline for online processing using CaImAn

Initialization

  • Get CNMFParams or similar object from server
  • Instantiate OnACID object using CNMFParams
  • Call OnACID.initialize_online()

Real-time processing

Overall need to break up and wrap existing OnACID.fit_online()

  • Wait for frame from server
  • For each given frame, run local method process_frame()
    • Adjust frame based on downsampling, normalizing
    • Motion correction
    • If requested, also update shapes for neuron finding
  • Return Estimates or similar object to data store

Benchmark

Memory Consumption

Suite2p

Instead of CaImAn, Suite2p could be used to process raw frames. This is demonstrated in batch_process.py.

Parameters

  • buffer_size=200 Number of frames to accumulate before starting Suite2p.
  • path='../output' Path to saved TIFF files.
  • max_workers=2 Maximum number of Suite2p instances to be running concurrently.

Description

Since Suite2p cannot perform an online analysis, we have to save raw frames into batches of size buffer_size.Each batch is stored as a stack in a TIFF file as specified in path. Afterwards, Suite2p is called. If Suite2p has not finished processing when a new batch arrives, a new process is spawned to run another instance of Suite2p until max_workers is reached.

Clone this wiki locally