Quick start

Once DeLTA is installed, you can already use it without writing any code, with its default configuration, through its command-line interface.

You can read the instructions to use the command-line interface by calling delta with the --help flag:

$ delta --help

usage: delta [-h] {run,train,evaluate,compare} ...

Deep Learning for Time-Lapse Analysis

positional arguments:
  {run,train,evaluate,compare}
                        Action to perform
    run                 Segment and track an experiment
    train               Train DeLTA's models on your dataset
    evaluate            Evaluate DeLTA's performance on your dataset
    compare             Compare two nc files (for debugging)

options:
  -h, --help            show this help message and exit

The options available for any of these commands are obtained in the same way, for example for run:

$ delta run --help

usage: delta run [-h] -c CONFIG -i INPUT [-o OUTPUT] [-C KEY=VALUE] [--positions POSITIONS]
                 [--frames FRAMES] [--progress] [--label-movie]

options:
  -h, --help            show this help message and exit
  -c, --config CONFIG   Configuration file. Can be either `2D`, `mothermachine`, or a path
                        to a previously saved custom config file.
  -i, --input INPUT     Input file or directory. Can include `{p}`, `{c}` and `{t}` as
                        placeholders for the position, channel, and frame number. Example
                        for micromanager:
                        `/path/to/folder/Pos{p}/img_channel{c}_position{p}_time{t}_z000.tif`
  -o, --output OUTPUT   Output directory (by default `delta_results` inside the input
                        directory)
  -C KEY=VALUE          Configuration option added on the fly for this run, for example
                        `min_cell_area=40`.
  --positions POSITIONS
                        Positions to process, ex.: 0-2,4,7-10 (default: all)
  --frames FRAMES       Range of frames to process, ex.: -150 (up to frame 150), 15- (from
                        frame 15), 15-30 (frames 15 to 30), 40 (just frame 40) (default:
                        all)
  --progress            Display progress bars.
  --label-movie         Label movie with cellids.

Here, the only two arguments required are CONFIG (use mothermachine if you are using one, 2D otherwise, or the path to a config file that you previously saved), and INPUT. If your images are saved as a single file, for example an nd2 or multi-page tiff file, just provide its name, but if they are saved as individual files inside a directory, then provide their naming pattern with the placeholders {p} for the position number, {c} for the channel number and {t} for the frame number, as needed (they are all optional).

Note

We refer to the term “position” for the (x, y) position of the microscope over the sample, “channel” for the imaging channel (the first must be bright field or phase contrast, and the rest are the fluorescence channels), and “frame” for the successive images taken of a given position as time progresses.

Example uses

Note

If you don’t have your own movies, you can download DeLTA’s demo movies with

>>> import delta
>>>
>>> # The paths returned might differ on your computer
>>> delta.assets.download_demo_movie("2D")
PosixPath('/home/virgile/.cache/delta/demo_movies/v3.0.0a6/unzipped/2D_demo/pos{p}cha{c}fra{t}.png')
>>> delta.assets.download_demo_movie("mothermachine")
PosixPath('/home/virgile/.cache/delta/demo_movies/v3.0.0a6/unzipped/mothermachine_demo/Pos{p}Chan{c}Frames{t}.png')

You can then launch DeLTA on these demo movies, specifying the -c and -i arguments with the mode (2D or mothermachine) and the path returned, as shown below.

• You recorded mothermachine images as a .nd2 file: you can just launch DeLTA on them with

$ delta run -c mothermachine -i my_images.nd2

• You did an agar pad experiment where each frame is an individual .tif image, all saved in the same directory images/ with names posXXX_channelXXX_tXXX.tif. Then, you can launch DeLTA on them with

$ delta run -c 2D -i images/pos{p}_channel{c}_t{t}.tif