(3->2) Add method to update Instances across views in RecordingSession

sleap/gui/app.py

@@ -1231,8 +1231,17 @@ def plotFrame(self, *args, **kwargs):
 
     def _after_plot_update(self, frame_idx):
         """Run after plot is updated, but stay on same frame."""
+
+        video = self.state["video"]
+
+        # Redraw trails
         overlay: TrackTrailOverlay = self.overlays["trails"]
-        overlay.redraw(self.state["video"], frame_idx)
+        overlay.redraw(video, frame_idx)
+
+        # Replot connected views for multi-camera projects
+        # TODO(LM): Use context.state["session"] in command instead (when implemented)
+        session = self.labels.get_session(video)
+        self.commands.triangulateSession(session=session)
 
     def _after_plot_change(self, player, frame_idx, selected_inst):
         """Called each time a new frame is drawn."""

sleap/gui/commands.py

@@ -35,14 +35,15 @@ class which inherits from `AppCommand` (or a more specialized class such as
 import traceback
 from enum import Enum
 from glob import glob
+from itertools import permutations, product
 from pathlib import Path, PurePath
-from typing import Callable, Dict, Iterator, List, Optional, Tuple, Type, Union
+from typing import Callable, Dict, Iterator, List, Optional, Tuple, Type, Union, cast
 
 import attr
 import cv2
 import numpy as np
-from qtpy import QtCore, QtWidgets, QtGui
-from qtpy.QtWidgets import QMessageBox, QProgressDialog
+from sleap_anipose import triangulate, reproject
+from qtpy import QtCore, QtGui, QtWidgets
 
 from sleap.gui.dialogs.delete import DeleteDialog
 from sleap.gui.dialogs.filedialog import FileDialog
@@ -53,7 +54,7 @@ class which inherits from `AppCommand` (or a more specialized class such as
 from sleap.gui.state import GuiState
 from sleap.gui.suggestions import VideoFrameSuggestions
 from sleap.instance import Instance, LabeledFrame, Point, PredictedInstance, Track
-from sleap.io.cameras import RecordingSession
+from sleap.io.cameras import Camcorder, InstanceGroup, FrameGroup, RecordingSession
 from sleap.io.convert import default_analysis_filename
 from sleap.io.dataset import Labels
 from sleap.io.format.adaptor import Adaptor
@@ -613,6 +614,20 @@ def generateSuggestions(self, params: Dict):
         """Generates suggestions using given params dictionary."""
         self.execute(GenerateSuggestions, **params)
 
+    def triangulateSession(
+        self,
+        session: Optional[RecordingSession] = None,
+        frame_idx: Optional[int] = None,
+        instance: Optional[Instance] = None,
+    ):
+        """Triangulates `Instance`s for selected views in a `RecordingSession`."""
+        self.execute(
+            TriangulateSession,
+            session=session,
+            frame_idx=frame_idx,
+            instance=instance,
+        )
+
     def openWebsite(self, url):
         """Open a website from URL using the native system browser."""
         self.execute(OpenWebsite, url=url)
@@ -1929,7 +1944,6 @@ class AddSession(EditCommand):
 
     @staticmethod
     def do_action(context: CommandContext, params: dict):
-
         camera_calibration = params["camera_calibration"]
         session = RecordingSession.load(filename=camera_calibration)
 
@@ -3369,6 +3383,143 @@ def do_action(cls, context: CommandContext, params: dict):
             context.labels.append(current_frame)
 
 
+class TriangulateSession(EditCommand):
+    topics = [UpdateTopic.frame, UpdateTopic.project_instances]
+
+    @classmethod
+    def do_action(cls, context: CommandContext, params: dict):
+        """Triangulate, reproject, and update instances in a session at a frame index.
+
+        Args:
+            context: The command context.
+            params: The command parameters.
+                session: The `RecordingSession` object to use. Default is current
+                    video's session.
+                frame_idx: The frame index to use. Default is current frame index.
+                instance: The `Instance` object to use. Default is current instance.
+                show_dialog: If True, then show a warning dialog. Default is True.
+                ask_again: If True, then ask for views/instances again. Default is False.
+        """
+
+        session: RecordingSession = (
+            params.get("session", None) or context.state["session"]
+        )
+        if session is None:
+            return
+
+        # Get `FrameGroup` for the current frame index
+        frame_idx: int = params.get("frame_idx", None)
+        if frame_idx is None:
+            frame_idx = context.state["frame_idx"]
+        frame_group: FrameGroup = session.frame_groups.get(frame_idx, None)
+        if frame_group is None:
+            return
+
+        # Get the `InstanceGroup` from `Instance` if any
+        instance = params.get("instance", None) or context.state["instance"]
+        instance_group = frame_group.get_instance_group(instance)
+
+        # If instance_group is None, then we will try to triangulate entire frame_group
+        instance_groups = (
+            [instance_group]
+            if instance_group is not None
+            else frame_group.instance_groups
+        )
+
+        # Retain instance groups that have enough views/instances for triangulation
+        instance_groups = TriangulateSession.has_enough_instances(
+            frame_group=frame_group,
+            instance_groups=instance_groups,
+            frame_idx=frame_idx,
+            instance=instance,
+        )
+        if instance_groups is None or len(instance_groups) == 0:
+            return  # Not enough instances for triangulation
+
+        # Get the `FrameGroup` of shape  M=include x T x N x 2
+        fg_tensor = frame_group.numpy(instance_groups=instance_groups, pred_as_nan=True)
+
+        # Add extra dimension for number of frames
+        frame_group_tensor = np.expand_dims(fg_tensor, axis=1)  # M=include x F=1 xTxNx2
+
+        # Triangulate to one 3D pose per instance
+        points_3d = triangulate(
+            p2d=frame_group_tensor,
+            calib=session.camera_cluster,
+            excluded_views=frame_group.excluded_views,
+        )  # F x T x N x 3
+
+        # Reproject onto all views
+        pts_reprojected = reproject(
+            points_3d,
+            calib=session.camera_cluster,
+            excluded_views=frame_group.excluded_views,
+        )  # M=include x F=1 x T x N x 2
+
+        # Sqeeze back to the original shape
+        points_reprojected = np.squeeze(pts_reprojected, axis=1)  # M=include x TxNx2
+
+        # Update or create/insert ("upsert") instance points
+        frame_group.upsert_points(
+            points=points_reprojected,
+            instance_groups=instance_groups,
+            exclude_complete=True,
+        )
+
+    @classmethod
+    def has_enough_instances(
+        cls,
+        frame_group: FrameGroup,
+        instance_groups: Optional[List[InstanceGroup]] = None,
+        frame_idx: Optional[int] = None,
+        instance: Optional[Instance] = None,
+    ) -> Optional[List[InstanceGroup]]:
+        """Filters out instance groups without enough instances for triangulation.
+
+        Args:
+            frame_group: The `FrameGroup` object to use.
+            instance_groups: A list of `InstanceGroup` objects to use. Default is None.
+            frame_idx: The frame index to use (only used in logging). Default is None.
+            instance: The `Instance` object to use (only used in logging). Default None.
+
+        Returns:
+            A list of `InstanceGroup` objects with enough instances for triangulation.
+        """
+
+        if instance is None:
+            instance = ""  # Just used for logging
+
+        if frame_idx is None:
+            frame_idx = ""  # Just used for logging
+
+        if instance_groups is None:
+            instance_groups = frame_group.instance_groups
+
+        if len(instance_groups) < 1:
+            logger.warning(
+                f"Require at least 1 instance group, but found "
+                f"{len(frame_group.instance_groups)} for frame group {frame_group} at "
+                f"frame {frame_idx}."
+                f"\nSkipping triangulation."
+            )
+            return None  # No instance groups found
+
+        # Assert that there are enough views and instances
+        instance_groups_to_tri = []
+        for instance_group in instance_groups:
+            instances = instance_group.get_instances(frame_group.cams_to_include)
+            if len(instances) < 2:
+                # Not enough instances
+                logger.warning(
+                    f"Not enough instances in {instance_group} for triangulation."
+                    f"\nSkipping instance group."
+                )
+                continue
+            instance_groups_to_tri.append(instance_group)
+
+        return instance_groups_to_tri  # `InstanceGroup`s with enough instances
+
+
 def open_website(url: str):
     """Open website in default browser.
 

sleap/gui/dialogs/delete.py

@@ -216,7 +216,7 @@ def _delete(self, lf_inst_list: List[Tuple[LabeledFrame, Instance]]):
         for lf, inst in lf_inst_list:
             self.context.labels.remove_instance(lf, inst, in_transaction=True)
             if not lf.instances:
-                self.context.labels.remove(lf)
+                self.context.labels.remove_frame(lf=lf, update_cache=False)
 
         # Update caches since we skipped doing this after each deletion
         self.context.labels.update_cache()

sleap/instance.py

@@ -17,21 +17,21 @@
   out of sync if the skeleton is manipulated.
 """
 
+import logging
 import math
-
-import numpy as np
-import cattr
-
 from copy import copy
-from typing import Dict, List, Optional, Union, Tuple, ForwardRef
+from typing import Dict, ForwardRef, List, Optional, Tuple, Union
 
+import attr
+import cattr
+import numpy as np
 from numpy.lib.recfunctions import structured_to_unstructured
 
 import sleap
-from sleap.skeleton import Skeleton, Node
 from sleap.io.video import Video
+from sleap.skeleton import Node, Skeleton
 
-import attr
+logger = logging.getLogger(__name__)
 
 
 class Point(np.record):
@@ -57,7 +57,6 @@ def __new__(
         visible: bool = True,
         complete: bool = False,
     ) -> "Point":
-
         # HACK: This is a crazy way to instantiate at new Point but I can't figure
         # out how recarray does it. So I just use it to make matrix of size 1 and
         # index in to get the np.record/Point
@@ -124,7 +123,6 @@ def __new__(
         complete: bool = False,
         score: float = 0.0,
     ) -> "PredictedPoint":
-
         # HACK: This is a crazy way to instantiate at new Point but I can't figure
         # out how recarray does it. So I just use it to make matrix of size 1 and
         # index in to get the np.record/Point
@@ -184,7 +182,6 @@ def __new__(
         aligned=False,
         order="C",
     ) -> "PointArray":
-
         dtype = subtype._record_type.dtype
 
         if dtype is not None:
@@ -445,12 +442,10 @@ def __attrs_post_init__(self):
         # If the user did not pass a points list initialize a point array for future
         # points.
         if self._points is None or len(self._points) == 0:
-
             # Initialize an empty point array that is the size of the skeleton.
             self._points = self._point_array_type.make_default(len(self.skeleton.nodes))
 
         else:
-
             if type(self._points) is dict:
                 parray = self._point_array_type.make_default(len(self.skeleton.nodes))
                 Instance._points_dict_to_array(self._points, parray, self.skeleton)
@@ -505,9 +500,11 @@ def _points_dict_to_array(
                 )
             try:
                 parray[skeleton.node_to_index(node)] = point
-                # parray[skeleton.node_to_index(node.name)] = point
             except:
-                pass
+                logger.debug(
+                    f"Could not set point for node {node} in {skeleton} "
+                    f"with point {point}"
+                )
 
     def _node_to_index(self, node: Union[str, Node]) -> int:
         """Helper method to get the index of a node from its name.
@@ -720,6 +717,45 @@ def points(self) -> Tuple[Point, ...]:
         self._fix_array()
         return tuple(point for point in self._points if not point.isnan())
 
+    def update_points(self, points: np.ndarray, exclude_complete: bool = False):
+        """Update the points in this instance from a numpy.
+
+        Args:
+            points: The new points to update to.
+            exclude_complete: Whether to update points where Point.complete is True
+        """
+        points_dict = dict()
+        for point_new, points_old, node_name in zip(
+            points, self._points, self.skeleton.node_names
+        ):
+
+            # Skip if new point is nan or old point is complete
+            if np.isnan(point_new).any() or (exclude_complete and points_old.complete):
+                continue
+
+            # Grab the x, y from the new point and visible, complete from the old point
+            x, y = point_new
+            visible = points_old.visible
+            complete = points_old.complete
+
+            # Create a new point and add to the dict
+            if type(self._points) == PredictedPointArray:
+                # TODO(LM): The point score is meant to rate the confidence of the
+                # prediction, but this method updates from triangulation.
+                score = points_old.score
+                point_obj = PredictedPoint(
+                    x=x, y=y, visible=visible, complete=complete, score=score
+                )
+            else:
+                point_obj = Point(x=x, y=y, visible=visible, complete=complete)
+
+            # Update the points dict
+            points_dict[node_name] = point_obj
+
+        # Update the points
+        if len(points_dict) > 0:
+            Instance._points_dict_to_array(points_dict, self._points, self.skeleton)
+
     def _fix_array(self):
         """Fix PointArray after nodes have been added or removed.
 
@@ -1199,7 +1235,6 @@ def make_instance_cattr() -> cattr.Converter:
     converter.register_unstructure_hook(PredictedPointArray, lambda x: None)
 
     def unstructure_instance(x: Instance):
-
         # Unstructure everything but the points array, nodes, and frame attribute
         d = {
             field.name: converter.unstructure(x.__getattribute__(field.name))
@@ -1380,7 +1415,9 @@ def find(
         Returns:
             List of instances.
         """
-        instances = self.instances
+        instances = sorted(
+            self.instances, key=lambda inst: isinstance(inst, PredictedInstance)
+        )  # Sort with PredictedInstances last
         if user:
             instances = list(filter(lambda inst: type(inst) == Instance, instances))
         if track != -1:  # use -1 since we want to accept None as possible value