core\refac: #68 handle symbolic tensor dims in loss

- handled sym tensor dimensions in tgce loss for allowing its usage as a
  keras sequence directly into fit methods

core/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 description = "Framework for handling image segmentation in the context of multiple annotators"
 name = "seg_tgce"
-version = "0.1.4.dev2"
+version = "0.1.4.dev3"
 readme = "README.md"
 authors = [{ name = "Brandon Lotero", email = "[email protected]" }]
 maintainers = [{ name = "Brandon Lotero", email = "[email protected]" }]
@@ -15,7 +15,7 @@ Issues = "https://github.com/blotero/seg_tgce/issues"
 
 [tool.poetry]
 name = "seg_tgce"
-version = "0.1.4.dev2"
+version = "0.1.4.dev3"
 authors = ["Brandon Lotero <[email protected]>"]
 description = "A package for the SEG TGCE project"
 readme = "README.md"

core/seg_tgce/loss/tgce.py

@@ -2,6 +2,7 @@
 from typing import Any
 
 import keras.backend as K
+from matplotlib.pylab import f
 import tensorflow as tf
 from keras.losses import Loss
 from tensorflow import cast
@@ -132,12 +133,21 @@ def call(self, y_true: tf.Tensor, y_pred: tf.Tensor) -> tf.Tensor:
         y_pred = cast(y_pred, TARGET_DATA_TYPE)
 
         y_pred = y_pred[..., : self.num_classes + self.num_annotators]  # type:ignore
-        y_true = tf.reshape(
-            y_true, (y_true.shape[:-1]) + (self.num_classes, self.num_annotators)
+
+        y_true_shape = tf.shape(y_true)
+
+        new_shape = tf.concat(
+            [y_true_shape[:-1], [self.num_classes, self.num_annotators]], axis=0
         )
+        y_true = tf.reshape(y_true, new_shape)
+
         lambda_r = y_pred[..., self.num_classes :]  # type:ignore
         y_pred_ = y_pred[..., : self.num_classes]
-        n_samples, width, height, _ = y_pred_.shape
+
+        n_samples = tf.shape(y_pred_)[0]
+        width = tf.shape(y_pred_)[1]
+        height = tf.shape(y_pred_)[2]
+
         y_pred_ = y_pred_[..., tf.newaxis]  # type:ignore
         y_pred_ = tf.repeat(y_pred_, repeats=[self.num_annotators], axis=-1)