keras-team/keras · error · ValueError

Input images must have 3 channels, but received images with

Error message

Input images must have 3 channels, but received images with {channels} channels.

What it means

RandomColorDegeneration converts images to grayscale internally, so compute_output_shape enforces exactly 3 channels (RGB). A channel count that is known and != 3 (1, 4, or more) fails model building with this ValueError.

Source

Thrown at keras/src/layers/preprocessing/image_preprocessing/random_color_degeneration.py:151

            {
                "factor": self.factor,
                "value_range": self.value_range,
                "seed": self.seed,
            }
        )
        return config

    def compute_output_shape(self, input_shape):
        if len(input_shape) not in (3, 4):
            raise ValueError(
                "Invalid images rank: expected rank 3 (single image) "
                "or rank 4 (batch of images). "
                f"Received: input_shape={input_shape}"
            )
        channels_axis = -1 if self.data_format == "channels_last" else -3
        channels = input_shape[channels_axis]
        if channels is not None and channels != 3:
            raise ValueError(
                "Input images must have 3 channels, but received images with "
                f"{channels} channels."
            )
        return input_shape


if RandomColorDegeneration.__doc__ is not None:
    RandomColorDegeneration.__doc__ = RandomColorDegeneration.__doc__.replace(
        "{{base_image_preprocessing_color_example}}",
        base_image_preprocessing_color_example.replace(
            "{LayerName}", "RandomColorDegeneration"
        ),
    )

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Convert inputs to RGB before the layer: tf.image.grayscale_to_rgb or np.repeat(x, 3, axis=-1) for 1-channel
2. Drop the alpha channel for RGBA: x[..., :3]
3. Apply color degeneration only on 3-channel branches

Example fix

# before
x = keras.Input(shape=(224, 224, 1))
x = RandomColorDegeneration(0.5)(x)
# after
x = keras.Input(shape=(224, 224, 3))
# or pre-convert: images = tf.image.grayscale_to_rgb(images)

c = images.shape[-1]
if c is not None and c != 3:
    images = np.repeat(images, 3 // c, axis=-1) if c == 1 else images[..., :3]

def is_rgb(x):
    c = x.shape[-1]
    return c is None or c == 3

When it happens

Trigger: keras.Input(shape=(224, 224, 1)) or (224, 224, 4) fed through RandomColorDegeneration and the model built/summarized.

Common situations: Grayscale datasets; RGBA images with alpha channel; medical imaging with multi-channel inputs.

Related errors

• Invalid images rank: expected rank 3 (single image) or rank
• `input_shape` must be a non-nested tuple or list of rank-1 w
• self._VALUE_RANGE_VALIDATION_ERROR + f"Received: value_range
• Expected the input image to be rank 3 or 4. Received inputs.
• Unknown activation function '{activation}' cannot be seriali

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/0034b2d19a10a179. Report an issue: GitHub.