qurator-spk/eynollah
1
0
Fork 0
mirror of https://github.com/qurator-spk/eynollah.git synced 2026-03-01 21:02:00 +01:00
Code Issues Projects Releases Wiki Activity

training.models: simplify CTC loss layer

Browse source
This commit is contained in:
Robert Sachunsky 2026-02-24 20:43:50 +01:00
parent 92fc2bd815
commit b399db3c00
1 changed files with 3 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

9
src/eynollah/training/models.py
View file

@ -29,6 +29,7 @@ from tensorflow.keras.layers import (
)
from tensorflow.keras.models import Model
from tensorflow.keras.regularizers import l2
from tensorflow.keras.backend import ctc_batch_cost
from ..patch_encoder import Patches, PatchEncoder
@ -45,10 +46,6 @@ MERGE_AXIS = -1
class CTCLayer(Layer):
def __init__(self, name=None):
super().__init__(name=name)
self.loss_fn = tf.keras.backend.ctc_batch_cost
def call(self, y_true, y_pred):
batch_len = tf.cast(tf.shape(y_true)[0], dtype="int64")
input_length = tf.cast(tf.shape(y_pred)[1], dtype="int64")
@ -56,7 +53,7 @@ class CTCLayer(Layer):
input_length = input_length * tf.ones(shape=(batch_len, 1), dtype="int64")
label_length = label_length * tf.ones(shape=(batch_len, 1), dtype="int64")
loss = self.loss_fn(y_true, y_pred, input_length, label_length)
loss = ctc_batch_cost(y_true, y_pred, input_length, label_length)
self.add_loss(loss)
# At test time, just return the computed predictions.
@ -505,6 +502,6 @@ def cnn_rnn_ocr_model(image_height=None, image_width=None, n_classes=None, max_s
# Add CTC layer for calculating CTC loss at each step.
output = CTCLayer(name="ctc_loss")(labels, out)
model = Model(inputs=[input_img, labels], outputs=output, name="handwriting_recognizer")
model = Model(inputs=(input_img, labels), outputs=output, name="handwriting_recognizer")
return model