Reduce tensorflow verbosity

requirements/common.txt

@@ -1,3 +1 @@
-tensorflow
 numpy
-matplotlib

requirements/plt.txt

@@ -0,0 +1,2 @@
+-r common.txt
+matplotlib

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = ccaptchas
-version = 0.0.1
+version = 0.1.2
 author = Daniil Fajnberg
 author_email = mail@daniil.fajnberg.de
 description = Character CAPTCHA Solver
@@ -17,14 +17,15 @@ classifiers =
 package_dir =
     = src
 packages = find:
-python_requires = >=3
+python_requires = >=3.9
 install_requires =
     numpy
-    matplotlib
 
 [options.extras_require]
 dev =
     coverage
+plt =
+    matplotlib
 
 [options.packages.find]
 where = src

src/ccaptchas/__main__.py

@@ -1,3 +1,5 @@
+import os
+import warnings
 from argparse import ArgumentParser
 from pathlib import Path
 from typing import Any, Sequence
@@ -156,4 +158,8 @@ def main() -> None:
 
 
 if __name__ == '__main__':
+    # Shut up Tensorflow:
+    os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
+    # Shut up numpy:
+    warnings.filterwarnings('ignore', 'elementwise comparison failed', FutureWarning, module='numpy')
     main()

src/ccaptchas/infer.py

@@ -11,7 +11,6 @@ from keras.api._v2.keras.backend import ctc_decode
 from .config import CONFIG
 from .preprocess import process_image, decode_label, find_image_files, get_lookup_table
 from .types import PathT, ImgT, Array
-from .visualize import plot_images
 
 
 def process_predictions(predictions: tf.Tensor) -> tf.Tensor:
@@ -53,7 +52,7 @@ def load_inference_model(model_dir: PathT) -> tuple[Model, StringLookup]:
 
 def predict_and_decode(images: Sequence[ImgT], model: Model, backward_lookup: StringLookup) -> tuple[Array, list[str]]:
     dataset = np.array([process_image(img) for img in images])
-    encoded_labels = process_predictions(model.predict(dataset))
+    encoded_labels = process_predictions(model.predict(dataset, verbose=0))
     return dataset, [decode_label(label, backward_lookup) for label in encoded_labels]
 
 
@@ -61,6 +60,7 @@ def load_and_infer(images: Sequence[ImgT], model_dir: PathT, plot_results: bool
     model, backward_lookup = load_inference_model(model_dir)
     images, labels = predict_and_decode(images, model, backward_lookup)
     if plot_results:
+        from .visualize import plot_images
         per_plot = 24
         for i in range(0, len(images), per_plot):
             plot_images(images[i:(i + per_plot)], labels=labels[i:(i + per_plot)])