diff --git a/frigate/config/logger.py b/frigate/config/logger.py
index 0e49f7422..a3eed23d0 100644
--- a/frigate/config/logger.py
+++ b/frigate/config/logger.py
@@ -31,6 +31,7 @@ class LoggerConfig(FrigateBaseModel):
             log_levels = {
                 "absl": LogLevel.error,
                 "httpx": LogLevel.error,
+                "tensorflow": LogLevel.error,
                 "werkzeug": LogLevel.error,
                 "ws4py": LogLevel.error,
                 **self.logs,
diff --git a/frigate/util/classification.py b/frigate/util/classification.py
index a8624870b..92da7c93e 100644
--- a/frigate/util/classification.py
+++ b/frigate/util/classification.py
@@ -1,7 +1,7 @@
 """Util for classification models."""
 
-import logging
 import os
+import sys
 
 import cv2
 import numpy as np
@@ -50,7 +50,13 @@ def train_classification_model(model_name: str) -> bool:
         ]
     )
 
-    tf.get_logger().setLevel(logging.ERROR)
+    # TF and Keras are very loud with logging
+    # we want to avoid these logs so we
+    # temporarily redirect stdout / stderr
+    original_stdout = sys.stdout
+    original_stderr = sys.stderr
+    sys.stdout = open(os.devnull, "w")
+    sys.stderr = open(os.devnull, "w")
 
     # Start with imagenet base model with 35% of channels in each layer
     base_model = MobileNetV2(
@@ -112,3 +118,7 @@ def train_classification_model(model_name: str) -> bool:
     # write model
     with open(os.path.join(model_dir, "model.tflite"), "wb") as f:
         f.write(tflite_model)
+
+    # restore original stdout / stderr
+    sys.stdout = original_stdout
+    sys.stderr = original_stderr