Creating Training Targets¶

In this step, we will prepare the target data necessary for training our model and predicting the coordinates of proteins within a tomogram.

We will use Copick to manage the filesystem, extract runIDs, and create spherical targets corresponding to the locations of proteins. Key tasks include:

• Generating Targets: For each tomogram, we extract particle coordinates and generate spherical targets based on these coordinates, and save the target data in OME Zarr format.

• Target dimensions are determined with an associated tomogram, (specified by the --tomo-uri parameter, e.g. wbp@10.0).

• Additional segmentations like organelles and membranes can be included as continuous targets by passing their name via --target — whether a name refers to particle picks or a continuous segmentation is auto-detected from the CoPick config, so no separate flag is needed.

The segmentations are saved under the query specified by the --target-uri flag (name:user_id/session_id).

Method 1: Automated Query¶

The simplest approach is to let Octopi automatically find all pickable objects from a specific annotation source.

When to Use¶

• ✅ Single annotation source: All picks come from one user/session
• ✅ Quick setup: Minimal parameter specification

Example Command¶

octopi create-targets \
    --config config.json \
    --picks-user-id data-portal --picks-session-id 0 \
    --tomo-uri wbp@10.0 \
    --target-uri targets:octopi/1

This command automatically finds every pickable object in the config and queries data-portal/0 for it — particle picks and continuous segmentations (like membrane) alike, since the type is auto-detected per object. If a particular object's annotations live under a different user/session than the rest (e.g. a membrane segmentation from a separate tool), list objects explicitly with --target instead (Method 2).

Method 2: Manual Specification¶

For more control, manually specify which protein types and annotation sources to include. Users can define a subset of pickable objects (from the CoPick configuration file) by specifying the name, and optionally the userID and sessionID. This allows for creating customizable training targets from varying submission sources.

When to Use¶

• ✅ Multiple annotation sources: Combining picks from different tools/users
• ✅ Selective training: Only specific protein types needed
• ✅ Quality control: Excluding low-quality annotations

Advanced Command¶

octopi create-targets \
    --config config.json \
    --target apoferritin --target beta-galactosidase:slabpick/1 \
    --target ribosome:pytom/0 --target virus-like-particle:pytom/0 \
    --target membrane:membrane-seg/1 \
    --tomo-uri wbp@10.0 \
    --target-uri targets:octopi/1

membrane here is a continuous segmentation rather than a particle pick set, but no separate flag is needed — Octopi looks it up in the CoPick config and detects its type automatically. Labels are assigned in the order the --target flags are given, so this mixed list keeps its true relative order.

Target Specification Formats¶

--target accepts the same query grammar regardless of whether it names a particle pick set or a continuous segmentation:

Check Target Quality¶

To validate your training targets, refer to our interactive notebook: Inspect Segmentation Targets

This notebook shows how to load segmentation targets and overlay targets on tomograms.

Example of training targets overlaid on tomogram slices. Different colors represent different protein classes.

Command Reference: octopi create-targets¶

Input Arguments¶

*Either --target OR both --picks-session-id and --picks-user-id must be specified.

Processing Parameters¶

Output Arguments¶