Skip to content

position_dlc_selection.py

DLCPosSelection

Bases: SpyglassMixin, Manual

Specify collection of upstream DLCCentroid and DLCOrientation entries to combine into a set of position information

Source code in src/spyglass/position/v1/position_dlc_selection.py 
29
30
31
32
33
34
35
36
37
38
39
@schema
class DLCPosSelection(SpyglassMixin, dj.Manual):
    """
    Specify collection of upstream DLCCentroid and DLCOrientation entries
    to combine into a set of position information
    """

    definition = """
    -> DLCCentroid.proj(dlc_si_cohort_centroid='dlc_si_cohort_selection_name', centroid_analysis_file_name='analysis_file_name')
    -> DLCOrientation.proj(dlc_si_cohort_orientation='dlc_si_cohort_selection_name', orientation_analysis_file_name='analysis_file_name')
    """

DLCPosV1

Bases: SpyglassMixin, Computed

Combines upstream DLCCentroid and DLCOrientation entries into a single entry with a single Analysis NWB file

Source code in src/spyglass/position/v1/position_dlc_selection.py 
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
@schema
class DLCPosV1(SpyglassMixin, dj.Computed):
    """
    Combines upstream DLCCentroid and DLCOrientation
    entries into a single entry with a single Analysis NWB file
    """

    definition = """
    -> DLCPosSelection
    ---
    -> AnalysisNwbfile
    position_object_id      : varchar(80)
    orientation_object_id   : varchar(80)
    velocity_object_id      : varchar(80)
    pose_eval_result        : longblob
    """

    def make(self, key):
        """Populate the table with the combined position information.

        1. Fetches position and orientation data from the DLCCentroid and
        DLCOrientation tables.
        2. Creates NWB objects for position, orientation, and velocity.
        3. Generates an AnalysisNwbfile and adds the NWB objects to it.
        4. Inserts the key into the table, and the PositionOutput Merge table.
        """
        orig_key = copy.deepcopy(key)
        # Add to Analysis NWB file
        AnalysisNwbfile()._creation_times["pre_create_time"] = time()
        key["pose_eval_result"] = self.evaluate_pose_estimation(key)

        pos_nwb = (DLCCentroid & key).fetch_nwb()[0]
        ori_nwb = (DLCOrientation & key).fetch_nwb()[0]

        pos_obj = pos_nwb["dlc_position"].spatial_series["position"]
        vel_obj = pos_nwb["dlc_velocity"].time_series["velocity"]
        vid_frame_obj = pos_nwb["dlc_velocity"].time_series["video_frame_ind"]
        ori_obj = ori_nwb["dlc_orientation"].spatial_series["orientation"]

        position = pynwb.behavior.Position()
        orientation = pynwb.behavior.CompassDirection()
        velocity = pynwb.behavior.BehavioralTimeSeries()

        position.create_spatial_series(
            name=pos_obj.name,
            timestamps=np.asarray(pos_obj.timestamps),
            conversion=pos_obj.conversion,
            data=np.asarray(pos_obj.data),
            reference_frame=pos_obj.reference_frame,
            comments=pos_obj.comments,
            description=pos_obj.description,
        )

        orientation.create_spatial_series(
            name=ori_obj.name,
            timestamps=np.asarray(ori_obj.timestamps),
            conversion=ori_obj.conversion,
            data=np.asarray(ori_obj.data),
            reference_frame=ori_obj.reference_frame,
            comments=ori_obj.comments,
            description=ori_obj.description,
        )

        velocity.create_timeseries(
            name=vel_obj.name,
            timestamps=np.asarray(vel_obj.timestamps),
            conversion=vel_obj.conversion,
            unit=vel_obj.unit,
            data=np.asarray(vel_obj.data),
            comments=vel_obj.comments,
            description=vel_obj.description,
        )

        velocity.create_timeseries(
            name=vid_frame_obj.name,
            timestamps=np.asarray(vid_frame_obj.timestamps),
            unit=vid_frame_obj.unit,
            data=np.asarray(vid_frame_obj.data),
            description=vid_frame_obj.description,
            comments=vid_frame_obj.comments,
        )

        # Add to Analysis NWB file
        analysis_file_name = AnalysisNwbfile().create(key["nwb_file_name"])
        key["analysis_file_name"] = analysis_file_name
        nwb_analysis_file = AnalysisNwbfile()

        key.update(
            {
                "analysis_file_name": analysis_file_name,
                "position_object_id": nwb_analysis_file.add_nwb_object(
                    analysis_file_name, position
                ),
                "orientation_object_id": nwb_analysis_file.add_nwb_object(
                    analysis_file_name, orientation
                ),
                "velocity_object_id": nwb_analysis_file.add_nwb_object(
                    analysis_file_name, velocity
                ),
            }
        )

        nwb_analysis_file.add(
            nwb_file_name=key["nwb_file_name"],
            analysis_file_name=analysis_file_name,
        )
        self.insert1(key)

        from ..position_merge import PositionOutput

        # TODO: The next line belongs in a merge table function
        PositionOutput._merge_insert(
            [orig_key],
            part_name=to_camel_case(self.table_name.split("__")[-1]),
            skip_duplicates=True,
        )
        AnalysisNwbfile().log(key, table=self.full_table_name)

    def fetch1_dataframe(self) -> pd.DataFrame:
        """Return the position data as a DataFrame."""
        nwb_data = self.fetch_nwb()[0]
        index = pd.Index(
            np.asarray(nwb_data["position"].get_spatial_series().timestamps),
            name="time",
        )
        COLUMNS = [
            "video_frame_ind",
            "position_x",
            "position_y",
            "orientation",
            "velocity_x",
            "velocity_y",
            "speed",
        ]
        return pd.DataFrame(
            np.concatenate(
                (
                    np.asarray(
                        nwb_data["velocity"]
                        .time_series["video_frame_ind"]
                        .data,
                        dtype=int,
                    )[:, np.newaxis],
                    np.asarray(nwb_data["position"].get_spatial_series().data),
                    np.asarray(
                        nwb_data["orientation"].get_spatial_series().data
                    )[:, np.newaxis],
                    np.asarray(
                        nwb_data["velocity"].time_series["velocity"].data
                    ),
                ),
                axis=1,
            ),
            columns=COLUMNS,
            index=index,
        )

    def fetch_nwb(self, **kwargs):
        """Fetch the NWB file."""
        attrs = [a for a in self.heading.names if not a == "pose_eval_result"]
        return super().fetch_nwb(*attrs, **kwargs)

    @classmethod
    def evaluate_pose_estimation(cls, key):
        """Evaluate the pose estimation."""
        likelihood_thresh = []

        valid_fields = DLCSmoothInterpCohort.BodyPart().heading.names
        centroid_key = {k: val for k, val in key.items() if k in valid_fields}
        centroid_key["dlc_si_cohort_selection_name"] = key[
            "dlc_si_cohort_centroid"
        ]
        centroid_bodyparts, centroid_si_params = (
            DLCSmoothInterpCohort.BodyPart & centroid_key
        ).fetch("bodypart", "dlc_si_params_name")

        orientation_key = centroid_key.copy()
        orientation_key["dlc_si_cohort_selection_name"] = key[
            "dlc_si_cohort_orientation"
        ]
        orientation_bodyparts, orientation_si_params = (
            DLCSmoothInterpCohort.BodyPart & orientation_key
        ).fetch("bodypart", "dlc_si_params_name")

        for param in np.unique(
            np.concatenate((centroid_si_params, orientation_si_params))
        ):
            likelihood_thresh.append(
                (
                    DLCSmoothInterpParams() & {"dlc_si_params_name": param}
                ).fetch1("params")["likelihood_thresh"]
            )

        if len(np.unique(likelihood_thresh)) > 1:
            raise ValueError("more than one likelihood threshold used")

        like_thresh = likelihood_thresh[0]
        bodyparts = np.unique([*centroid_bodyparts, *orientation_bodyparts])
        fields = DLCPoseEstimation.BodyPart.heading.names
        pose_estimation_key = {k: v for k, v in key.items() if k in fields}
        pose_estimation_df = pd.concat(
            {
                bodypart: (
                    DLCPoseEstimation.BodyPart()
                    & {**pose_estimation_key, **{"bodypart": bodypart}}
                ).fetch1_dataframe()
                for bodypart in bodyparts.tolist()
            },
            axis=1,
        )
        df_filter = {
            bodypart: pose_estimation_df[bodypart]["likelihood"] < like_thresh
            for bodypart in bodyparts
            if bodypart in pose_estimation_df.columns
        }
        sub_thresh_percent_dict = {
            bodypart: (
                len(
                    np.where(
                        ~np.isnan(
                            pose_estimation_df[bodypart]["likelihood"].where(
                                df_filter[bodypart]
                            )
                        )
                    )[0]
                )
                / len(pose_estimation_df)
            )
            * 100
            for bodypart in bodyparts
        }
        return sub_thresh_percent_dict

make(key)

Populate the table with the combined position information.

  1. Fetches position and orientation data from the DLCCentroid and DLCOrientation tables.
  2. Creates NWB objects for position, orientation, and velocity.
  3. Generates an AnalysisNwbfile and adds the NWB objects to it.
  4. Inserts the key into the table, and the PositionOutput Merge table.
Source code in src/spyglass/position/v1/position_dlc_selection.py 
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
def make(self, key):
    """Populate the table with the combined position information.

    1. Fetches position and orientation data from the DLCCentroid and
    DLCOrientation tables.
    2. Creates NWB objects for position, orientation, and velocity.
    3. Generates an AnalysisNwbfile and adds the NWB objects to it.
    4. Inserts the key into the table, and the PositionOutput Merge table.
    """
    orig_key = copy.deepcopy(key)
    # Add to Analysis NWB file
    AnalysisNwbfile()._creation_times["pre_create_time"] = time()
    key["pose_eval_result"] = self.evaluate_pose_estimation(key)

    pos_nwb = (DLCCentroid & key).fetch_nwb()[0]
    ori_nwb = (DLCOrientation & key).fetch_nwb()[0]

    pos_obj = pos_nwb["dlc_position"].spatial_series["position"]
    vel_obj = pos_nwb["dlc_velocity"].time_series["velocity"]
    vid_frame_obj = pos_nwb["dlc_velocity"].time_series["video_frame_ind"]
    ori_obj = ori_nwb["dlc_orientation"].spatial_series["orientation"]

    position = pynwb.behavior.Position()
    orientation = pynwb.behavior.CompassDirection()
    velocity = pynwb.behavior.BehavioralTimeSeries()

    position.create_spatial_series(
        name=pos_obj.name,
        timestamps=np.asarray(pos_obj.timestamps),
        conversion=pos_obj.conversion,
        data=np.asarray(pos_obj.data),
        reference_frame=pos_obj.reference_frame,
        comments=pos_obj.comments,
        description=pos_obj.description,
    )

    orientation.create_spatial_series(
        name=ori_obj.name,
        timestamps=np.asarray(ori_obj.timestamps),
        conversion=ori_obj.conversion,
        data=np.asarray(ori_obj.data),
        reference_frame=ori_obj.reference_frame,
        comments=ori_obj.comments,
        description=ori_obj.description,
    )

    velocity.create_timeseries(
        name=vel_obj.name,
        timestamps=np.asarray(vel_obj.timestamps),
        conversion=vel_obj.conversion,
        unit=vel_obj.unit,
        data=np.asarray(vel_obj.data),
        comments=vel_obj.comments,
        description=vel_obj.description,
    )

    velocity.create_timeseries(
        name=vid_frame_obj.name,
        timestamps=np.asarray(vid_frame_obj.timestamps),
        unit=vid_frame_obj.unit,
        data=np.asarray(vid_frame_obj.data),
        description=vid_frame_obj.description,
        comments=vid_frame_obj.comments,
    )

    # Add to Analysis NWB file
    analysis_file_name = AnalysisNwbfile().create(key["nwb_file_name"])
    key["analysis_file_name"] = analysis_file_name
    nwb_analysis_file = AnalysisNwbfile()

    key.update(
        {
            "analysis_file_name": analysis_file_name,
            "position_object_id": nwb_analysis_file.add_nwb_object(
                analysis_file_name, position
            ),
            "orientation_object_id": nwb_analysis_file.add_nwb_object(
                analysis_file_name, orientation
            ),
            "velocity_object_id": nwb_analysis_file.add_nwb_object(
                analysis_file_name, velocity
            ),
        }
    )

    nwb_analysis_file.add(
        nwb_file_name=key["nwb_file_name"],
        analysis_file_name=analysis_file_name,
    )
    self.insert1(key)

    from ..position_merge import PositionOutput

    # TODO: The next line belongs in a merge table function
    PositionOutput._merge_insert(
        [orig_key],
        part_name=to_camel_case(self.table_name.split("__")[-1]),
        skip_duplicates=True,
    )
    AnalysisNwbfile().log(key, table=self.full_table_name)

fetch1_dataframe()

Return the position data as a DataFrame.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
def fetch1_dataframe(self) -> pd.DataFrame:
    """Return the position data as a DataFrame."""
    nwb_data = self.fetch_nwb()[0]
    index = pd.Index(
        np.asarray(nwb_data["position"].get_spatial_series().timestamps),
        name="time",
    )
    COLUMNS = [
        "video_frame_ind",
        "position_x",
        "position_y",
        "orientation",
        "velocity_x",
        "velocity_y",
        "speed",
    ]
    return pd.DataFrame(
        np.concatenate(
            (
                np.asarray(
                    nwb_data["velocity"]
                    .time_series["video_frame_ind"]
                    .data,
                    dtype=int,
                )[:, np.newaxis],
                np.asarray(nwb_data["position"].get_spatial_series().data),
                np.asarray(
                    nwb_data["orientation"].get_spatial_series().data
                )[:, np.newaxis],
                np.asarray(
                    nwb_data["velocity"].time_series["velocity"].data
                ),
            ),
            axis=1,
        ),
        columns=COLUMNS,
        index=index,
    )

fetch_nwb(**kwargs)

Fetch the NWB file.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
199
200
201
202
def fetch_nwb(self, **kwargs):
    """Fetch the NWB file."""
    attrs = [a for a in self.heading.names if not a == "pose_eval_result"]
    return super().fetch_nwb(*attrs, **kwargs)

evaluate_pose_estimation(key) classmethod

Evaluate the pose estimation.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
@classmethod
def evaluate_pose_estimation(cls, key):
    """Evaluate the pose estimation."""
    likelihood_thresh = []

    valid_fields = DLCSmoothInterpCohort.BodyPart().heading.names
    centroid_key = {k: val for k, val in key.items() if k in valid_fields}
    centroid_key["dlc_si_cohort_selection_name"] = key[
        "dlc_si_cohort_centroid"
    ]
    centroid_bodyparts, centroid_si_params = (
        DLCSmoothInterpCohort.BodyPart & centroid_key
    ).fetch("bodypart", "dlc_si_params_name")

    orientation_key = centroid_key.copy()
    orientation_key["dlc_si_cohort_selection_name"] = key[
        "dlc_si_cohort_orientation"
    ]
    orientation_bodyparts, orientation_si_params = (
        DLCSmoothInterpCohort.BodyPart & orientation_key
    ).fetch("bodypart", "dlc_si_params_name")

    for param in np.unique(
        np.concatenate((centroid_si_params, orientation_si_params))
    ):
        likelihood_thresh.append(
            (
                DLCSmoothInterpParams() & {"dlc_si_params_name": param}
            ).fetch1("params")["likelihood_thresh"]
        )

    if len(np.unique(likelihood_thresh)) > 1:
        raise ValueError("more than one likelihood threshold used")

    like_thresh = likelihood_thresh[0]
    bodyparts = np.unique([*centroid_bodyparts, *orientation_bodyparts])
    fields = DLCPoseEstimation.BodyPart.heading.names
    pose_estimation_key = {k: v for k, v in key.items() if k in fields}
    pose_estimation_df = pd.concat(
        {
            bodypart: (
                DLCPoseEstimation.BodyPart()
                & {**pose_estimation_key, **{"bodypart": bodypart}}
            ).fetch1_dataframe()
            for bodypart in bodyparts.tolist()
        },
        axis=1,
    )
    df_filter = {
        bodypart: pose_estimation_df[bodypart]["likelihood"] < like_thresh
        for bodypart in bodyparts
        if bodypart in pose_estimation_df.columns
    }
    sub_thresh_percent_dict = {
        bodypart: (
            len(
                np.where(
                    ~np.isnan(
                        pose_estimation_df[bodypart]["likelihood"].where(
                            df_filter[bodypart]
                        )
                    )
                )[0]
            )
            / len(pose_estimation_df)
        )
        * 100
        for bodypart in bodyparts
    }
    return sub_thresh_percent_dict

DLCPosVideoParams

Bases: SpyglassMixin, Manual

Source code in src/spyglass/position/v1/position_dlc_selection.py 
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
@schema
class DLCPosVideoParams(SpyglassMixin, dj.Manual):
    definition = """
    dlc_pos_video_params_name : varchar(50)
    ---
    params : blob
    """

    @classmethod
    def insert_default(cls):
        """Insert the default parameters."""
        params = {
            "percent_frames": 1,
            "incl_likelihood": True,
            "video_params": {
                "arrow_radius": 20,
                "circle_radius": 6,
            },
        }
        cls.insert1(
            {"dlc_pos_video_params_name": "default", "params": params},
            skip_duplicates=True,
        )

    @classmethod
    def get_default(cls):
        """Return the default parameters."""
        query = cls & {"dlc_pos_video_params_name": "default"}
        if not len(query) > 0:
            cls().insert_default()
            default = (cls & {"dlc_pos_video_params_name": "default"}).fetch1()
        else:
            default = query.fetch1()
        return default

insert_default() classmethod

Insert the default parameters.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
@classmethod
def insert_default(cls):
    """Insert the default parameters."""
    params = {
        "percent_frames": 1,
        "incl_likelihood": True,
        "video_params": {
            "arrow_radius": 20,
            "circle_radius": 6,
        },
    }
    cls.insert1(
        {"dlc_pos_video_params_name": "default", "params": params},
        skip_duplicates=True,
    )

get_default() classmethod

Return the default parameters.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
300
301
302
303
304
305
306
307
308
309
@classmethod
def get_default(cls):
    """Return the default parameters."""
    query = cls & {"dlc_pos_video_params_name": "default"}
    if not len(query) > 0:
        cls().insert_default()
        default = (cls & {"dlc_pos_video_params_name": "default"}).fetch1()
    else:
        default = query.fetch1()
    return default

DLCPosVideo

Bases: SpyglassMixin, Computed

Creates a video of the computed head position and orientation as well as the original LED positions overlaid on the video of the animal.

Use for debugging the effect of position extraction parameters.

Source code in src/spyglass/position/v1/position_dlc_selection.py 
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
@schema
class DLCPosVideo(SpyglassMixin, dj.Computed):
    """Creates a video of the computed head position and orientation as well as
    the original LED positions overlaid on the video of the animal.

    Use for debugging the effect of position extraction parameters."""

    definition = """
    -> DLCPosVideoSelection
    ---
    """

    def make(self, key):
        """Populate the DLCPosVideo table.

        1. Fetches parameters from the DLCPosVideoParams table.
        2. Fetches position interval name from epoch name.
        3. Fetches pose estimation data and video information.
        4. Fetches centroid and likelihood data for each bodypart.
        5. Calls make_video to create the video with the above data.
        """
        M_TO_CM = 100

        params = (DLCPosVideoParams & key).fetch1("params")

        interval_name = convert_epoch_interval_name_to_position_interval_name(
            {
                "nwb_file_name": key["nwb_file_name"],
                "epoch": key["epoch"],
            },
            populate_missing=False,
        )
        epoch = (
            int(interval_name.replace("pos ", "").replace(" valid times", ""))
            + 1
        )
        pose_est_key = {
            "nwb_file_name": key["nwb_file_name"],
            "epoch": epoch,
            "dlc_model_name": key["dlc_model_name"],
            "dlc_model_params_name": key["dlc_model_params_name"],
        }

        pose_estimation_params, video_filename, output_dir, meters_per_pixel = (
            DLCPoseEstimationSelection * DLCPoseEstimation & pose_est_key
        ).fetch1(
            "pose_estimation_params",
            "video_path",
            "pose_estimation_output_dir",
            "meters_per_pixel",
        )

        logger.info(f"video filename: {video_filename}")
        logger.info("Loading position data...")

        v1_key = {k: v for k, v in key.items() if k in DLCPosV1.primary_key}
        pos_info_df = (
            DLCPosV1() & {"epoch": epoch, **v1_key}
        ).fetch1_dataframe()
        pos_est_df = pd.concat(
            {
                bodypart: (
                    DLCPoseEstimation.BodyPart()
                    & {**pose_est_key, **{"bodypart": bodypart}}
                ).fetch1_dataframe()
                for bodypart in (DLCSmoothInterpCohort.BodyPart & pose_est_key)
                .fetch("bodypart")
                .tolist()
            },
            axis=1,
        )
        if not len(pos_est_df) == len(pos_info_df):
            raise ValueError(
                "Dataframes are not the same length\n"
                + f"\tPose estim   :  {len(pos_est_df)}\n"
                + f"\tPosition info: {len(pos_info_df)}"
            )

        output_video_filename = (
            key["nwb_file_name"].replace(".nwb", "")
            + f"_{epoch:02d}_"
            + f'{key["dlc_si_cohort_centroid"]}_'
            + f'{key["dlc_centroid_params_name"]}'
            + f'{key["dlc_orientation_params_name"]}.mp4'
        )
        if Path(output_dir).exists():
            output_video_filename = Path(output_dir) / output_video_filename

        idx = pd.IndexSlice
        video_frame_inds = pos_info_df["video_frame_ind"].astype(int).to_numpy()
        centroids = {
            bodypart: pos_est_df.loc[:, idx[bodypart, ("x", "y")]].to_numpy()
            for bodypart in pos_est_df.columns.levels[0]
        }
        likelihoods = (
            {
                bodypart: pos_est_df.loc[
                    :, idx[bodypart, ("likelihood")]
                ].to_numpy()
                for bodypart in pos_est_df.columns.levels[0]
            }
            if params.get("incl_likelihood")
            else None
        )
        frames = params.get("frames", None)

        make_video(
            video_filename=video_filename,
            video_frame_inds=video_frame_inds,
            position_mean={
                "DLC": np.asarray(pos_info_df[["position_x", "position_y"]])
            },
            orientation_mean={"DLC": np.asarray(pos_info_df[["orientation"]])},
            centroids=centroids,
            likelihoods=likelihoods,
            position_time=np.asarray(pos_info_df.index),
            processor=params.get("processor", "opencv"),
            frames=np.arange(frames[0], frames[1]) if frames else None,
            percent_frames=params.get("percent_frames", None),
            output_video_filename=output_video_filename,
            cm_to_pixels=meters_per_pixel * M_TO_CM,
            crop=pose_estimation_params.get("cropping"),
            **params.get("video_params", {}),
        )
        self.insert1(key)

make(key)

Populate the DLCPosVideo table.

  1. Fetches parameters from the DLCPosVideoParams table.
  2. Fetches position interval name from epoch name.
  3. Fetches pose estimation data and video information.
  4. Fetches centroid and likelihood data for each bodypart.
  5. Calls make_video to create the video with the above data.
Source code in src/spyglass/position/v1/position_dlc_selection.py 
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
def make(self, key):
    """Populate the DLCPosVideo table.

    1. Fetches parameters from the DLCPosVideoParams table.
    2. Fetches position interval name from epoch name.
    3. Fetches pose estimation data and video information.
    4. Fetches centroid and likelihood data for each bodypart.
    5. Calls make_video to create the video with the above data.
    """
    M_TO_CM = 100

    params = (DLCPosVideoParams & key).fetch1("params")

    interval_name = convert_epoch_interval_name_to_position_interval_name(
        {
            "nwb_file_name": key["nwb_file_name"],
            "epoch": key["epoch"],
        },
        populate_missing=False,
    )
    epoch = (
        int(interval_name.replace("pos ", "").replace(" valid times", ""))
        + 1
    )
    pose_est_key = {
        "nwb_file_name": key["nwb_file_name"],
        "epoch": epoch,
        "dlc_model_name": key["dlc_model_name"],
        "dlc_model_params_name": key["dlc_model_params_name"],
    }

    pose_estimation_params, video_filename, output_dir, meters_per_pixel = (
        DLCPoseEstimationSelection * DLCPoseEstimation & pose_est_key
    ).fetch1(
        "pose_estimation_params",
        "video_path",
        "pose_estimation_output_dir",
        "meters_per_pixel",
    )

    logger.info(f"video filename: {video_filename}")
    logger.info("Loading position data...")

    v1_key = {k: v for k, v in key.items() if k in DLCPosV1.primary_key}
    pos_info_df = (
        DLCPosV1() & {"epoch": epoch, **v1_key}
    ).fetch1_dataframe()
    pos_est_df = pd.concat(
        {
            bodypart: (
                DLCPoseEstimation.BodyPart()
                & {**pose_est_key, **{"bodypart": bodypart}}
            ).fetch1_dataframe()
            for bodypart in (DLCSmoothInterpCohort.BodyPart & pose_est_key)
            .fetch("bodypart")
            .tolist()
        },
        axis=1,
    )
    if not len(pos_est_df) == len(pos_info_df):
        raise ValueError(
            "Dataframes are not the same length\n"
            + f"\tPose estim   :  {len(pos_est_df)}\n"
            + f"\tPosition info: {len(pos_info_df)}"
        )

    output_video_filename = (
        key["nwb_file_name"].replace(".nwb", "")
        + f"_{epoch:02d}_"
        + f'{key["dlc_si_cohort_centroid"]}_'
        + f'{key["dlc_centroid_params_name"]}'
        + f'{key["dlc_orientation_params_name"]}.mp4'
    )
    if Path(output_dir).exists():
        output_video_filename = Path(output_dir) / output_video_filename

    idx = pd.IndexSlice
    video_frame_inds = pos_info_df["video_frame_ind"].astype(int).to_numpy()
    centroids = {
        bodypart: pos_est_df.loc[:, idx[bodypart, ("x", "y")]].to_numpy()
        for bodypart in pos_est_df.columns.levels[0]
    }
    likelihoods = (
        {
            bodypart: pos_est_df.loc[
                :, idx[bodypart, ("likelihood")]
            ].to_numpy()
            for bodypart in pos_est_df.columns.levels[0]
        }
        if params.get("incl_likelihood")
        else None
    )
    frames = params.get("frames", None)

    make_video(
        video_filename=video_filename,
        video_frame_inds=video_frame_inds,
        position_mean={
            "DLC": np.asarray(pos_info_df[["position_x", "position_y"]])
        },
        orientation_mean={"DLC": np.asarray(pos_info_df[["orientation"]])},
        centroids=centroids,
        likelihoods=likelihoods,
        position_time=np.asarray(pos_info_df.index),
        processor=params.get("processor", "opencv"),
        frames=np.arange(frames[0], frames[1]) if frames else None,
        percent_frames=params.get("percent_frames", None),
        output_video_filename=output_video_filename,
        cm_to_pixels=meters_per_pixel * M_TO_CM,
        crop=pose_estimation_params.get("cropping"),
        **params.get("video_params", {}),
    )
    self.insert1(key)