From 699cb033303e1152717400518ca529cbe67b9925 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Fran=C3=A7ois=20Laurent?= <francois.laurent@posteo.net>
Date: Tue, 14 Mar 2023 23:42:25 +0100
Subject: [PATCH] full_label_list factored into encoding_label_list and
decoding_label_list
---
src/taggingbackends/data/labels.py | 34 ++++++++++++++++++++++++------
1 file changed, 27 insertions(+), 7 deletions(-)
diff --git a/src/taggingbackends/data/labels.py b/src/taggingbackends/data/labels.py
index c9301e2..a962b57 100644
--- a/src/taggingbackends/data/labels.py
+++ b/src/taggingbackends/data/labels.py
@@ -66,7 +66,9 @@ because label colors are also stored in the `labelspec` attribute.
With taggingbackends==0.9, a related attribute was introduced:
`secondarylabelspec`. To get a unique array of indexable labels, both primary
-and secondary labels in a same array, use `full_label_list` instead.
+and secondary labels in a same array, use either `encoding_label_list` or
+`decoding_label_list` instead. The two properties typically equal, unless
+attribute `decodingspec` is defined.
"""
class Labels:
@@ -267,6 +269,7 @@ class Labels:
self.labels, self.metadata = new_self.labels, new_self.metadata
self.labelspec, self.units = new_self.labelspec, new_self.units
self.secondarylabelspec = new_self.secondarylabelspec
+ # no `decodingspec` in label files
self.tracking = new_self.tracking
return self
@@ -309,6 +312,7 @@ class Labels:
self.units = data.get("units", {})
self.labelspec = data.get("labels", {})
self.secondarylabelspec = data.get("secondarylabels", [])
+ self.decodingspec = None # ensure it is not set
self._tracking = data.get("dependencies", [])
if isinstance(self._tracking, dict):
self._tracking = [self._tracking]
@@ -325,7 +329,7 @@ class Labels:
List of str: all different labels including primary and secondary labels.
"""
@property
- def full_label_list(self):
+ def encoding_label_list(self):
if isinstance(self.labelspec, dict):
labelset = self.labelspec['names']
else:
@@ -334,6 +338,18 @@ class Labels:
labelset = labelset + self.secondarylabelspec
return labelset
+ """
+ List of str: all different labels including primary and secondary labels.
+ """
+ @property
+ def decoding_label_list(self):
+ labelspec = self.decodingspec
+ if labelspec is None:
+ return self.encoding_label_list
+ if self.secondarylabelspec:
+ labelset = labelset + self.secondarylabelspec
+ return labelset
+
"""
Encode the text labels as indices (`int` or `list` of `int`).
@@ -348,7 +364,7 @@ class Labels:
elif isinstance(label, dict):
encoded = {t: self.encode(l) for t, l in label.items()}
else:
- labelset = self.full_label_list
+ labelset = self.encoding_label_list
if isinstance(label, str):
encoded = labelset.index(label) + 1
elif isinstance(label, int):
@@ -373,6 +389,13 @@ class Labels:
defined by the tagger used to generate the label file. In the case of
decoding a label file, `labelspec` only should be defined in the `Labels`
object.
+
+ To forget about these considerations, call `load_model_config` first if
+ processing the output of a tagger, or do not call that method otherwise.
+
+ MaggotUBA's `predict_model` does not call `decode` but directly indexes into
+ `behavior_labels`. As a consequence, the above subtleties do not apply to
+ MaggotUBA.
"""
def decode(self, label=None):
if label is None:
@@ -382,10 +405,7 @@ class Labels:
elif isinstance(label, dict):
decoded = {t: self.decode(l) for t, l in label.items()}
else:
- if self.decodingspec is None:
- labelset = self.full_label_list
- else:
- labelset = self.decodingspec
+ labelset = self.decoding_label_list
if isinstance(label, int):
decoded = labelset[label-1]
elif isinstance(label, str):
--
GitLab