deploy: 3c7f5af

develop/_sources/api.rst

@@ -64,6 +64,12 @@ Developer API
    sleap.nn.training
    sleap.nn.utils
    sleap.nn.viz
+   sleap.nn.config.data
+   sleap.nn.config.model
+   sleap.nn.config.optimization
+   sleap.nn.config.outputs
+   sleap.nn.config.training_job
+   sleap.nn.config.utils
    sleap.nn.architectures.common
    sleap.nn.architectures.encoder_decoder
    sleap.nn.architectures.hourglass
@@ -75,12 +81,6 @@ Developer API
    sleap.nn.architectures.upsampling
    sleap.nn.tracker.components
    sleap.nn.tracker.kalman
-   sleap.nn.config.data
-   sleap.nn.config.model
-   sleap.nn.config.optimization
-   sleap.nn.config.outputs
-   sleap.nn.config.training_job
-   sleap.nn.config.utils
    sleap.nn.data.augmentation
    sleap.nn.data.confidence_maps
    sleap.nn.data.dataset_ops

develop/_sources/guides/cli.md

@@ -207,7 +207,7 @@ optional arguments:
   --tracking.clean_iou_threshold TRACKING.CLEAN_IOU_THRESHOLD
                         IOU to use when culling instances *after* tracking. (default: 0)
   --tracking.similarity TRACKING.SIMILARITY
-                        Options: instance, centroid, iou (default: instance)
+                        Options: instance, normalized_instance, object_keypoint, centroid, iou (default: instance)
   --tracking.match TRACKING.MATCH
                         Options: hungarian, greedy (default: greedy)
   --tracking.robust TRACKING.ROBUST

develop/_sources/guides/proofreading.md

@@ -50,6 +50,8 @@ There are currently three methods for matching instances in frame N against thes
 - “**centroid**” measures similarity by the distance between the instance centroids
 - “**iou**” measures similarity by the intersection/overlap of the instance bounding boxes
 - “**instance**” measures similarity by looking at the distances between corresponding nodes in the instances, normalized by the number of valid nodes in the candidate instance.
+- “**normalized_instance**” measures similarity by looking at the distances between corresponding nodes in the instances, normalized by the number of valid nodes in the candidate instance and the keypoints normalized by the image size.
+- “**object_keypoint**” measures similarity by measuring the distance between each keypoints from a reference instance and a query instance, takes the exp(-d**2), sum for all the keypoints and divide by the number of visible keypoints in the reference instance.
 
 Once SLEAP has measured the similarity between all the candidates and the instances in frame N, you need to choose a way to pair them up. You can do this either by picking the best match, and the picking the best remaining match for each remaining instance in turn—this is “**greedy**” matching—or you can find the way of matching identities which minimizes the total cost (or: maximizes the total similarity)—this is “**Hungarian**” matching.
 

develop/api.html

@@ -473,6 +473,24 @@ <h1>Developer API</h1>
 <tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.viz.html#module-sleap.nn.viz" title="sleap.nn.viz"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.viz</span></code></a></p></td>
 <td><p>Visualization and plotting utilities.</p></td>
 </tr>
+<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.data.html#module-sleap.nn.config.data" title="sleap.nn.config.data"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.data</span></code></a></p></td>
+<td><p></p></td>
+</tr>
+<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.model.html#module-sleap.nn.config.model" title="sleap.nn.config.model"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.model</span></code></a></p></td>
+<td><p></p></td>
+</tr>
+<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.optimization.html#module-sleap.nn.config.optimization" title="sleap.nn.config.optimization"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.optimization</span></code></a></p></td>
+<td><p></p></td>
+</tr>
+<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.outputs.html#module-sleap.nn.config.outputs" title="sleap.nn.config.outputs"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.outputs</span></code></a></p></td>
+<td><p></p></td>
+</tr>
+<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.training_job.html#module-sleap.nn.config.training_job" title="sleap.nn.config.training_job"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.training_job</span></code></a></p></td>
+<td><p>Serializable configuration classes for specifying all training job parameters.</p></td>
+</tr>
+<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.utils.html#module-sleap.nn.config.utils" title="sleap.nn.config.utils"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.utils</span></code></a></p></td>
+<td><p>Utilities for config building and validation.</p></td>
+</tr>
 <tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.architectures.common.html#module-sleap.nn.architectures.common" title="sleap.nn.architectures.common"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.architectures.common</span></code></a></p></td>
 <td><p>Common utilities for architecture and model building.</p></td>
 </tr>
@@ -506,24 +524,6 @@ <h1>Developer API</h1>
 <tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.tracker.kalman.html#module-sleap.nn.tracker.kalman" title="sleap.nn.tracker.kalman"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.tracker.kalman</span></code></a></p></td>
 <td><p>Module to use Kalman filters for tracking instance identities.</p></td>
 </tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.data.html#module-sleap.nn.config.data" title="sleap.nn.config.data"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.data</span></code></a></p></td>
-<td><p></p></td>
-</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.model.html#module-sleap.nn.config.model" title="sleap.nn.config.model"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.model</span></code></a></p></td>
-<td><p></p></td>
-</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.optimization.html#module-sleap.nn.config.optimization" title="sleap.nn.config.optimization"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.optimization</span></code></a></p></td>
-<td><p></p></td>
-</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.outputs.html#module-sleap.nn.config.outputs" title="sleap.nn.config.outputs"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.outputs</span></code></a></p></td>
-<td><p></p></td>
-</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.config.training_job.html#module-sleap.nn.config.training_job" title="sleap.nn.config.training_job"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.training_job</span></code></a></p></td>
-<td><p>Serializable configuration classes for specifying all training job parameters.</p></td>
-</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="api/sleap.nn.config.utils.html#module-sleap.nn.config.utils" title="sleap.nn.config.utils"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.config.utils</span></code></a></p></td>
-<td><p>Utilities for config building and validation.</p></td>
-</tr>
 <tr class="row-odd"><td><p><a class="reference internal" href="api/sleap.nn.data.augmentation.html#module-sleap.nn.data.augmentation" title="sleap.nn.data.augmentation"><code class="xref py py-obj docutils literal notranslate"><span class="pre">sleap.nn.data.augmentation</span></code></a></p></td>
 <td><p>Transformers for applying data augmentation.</p></td>
 </tr>