keras-team/keras · error · KeyError

There are unsupported keys in `bounding_boxes`: {list(extra_

Error message

There are unsupported keys in `bounding_boxes`: {list(extra_keys)}. Only `boxes` and `labels` are supported.

What it means

MaxNumBoundingBoxes.compute_output_shape only accepts 'boxes' and 'labels' keys inside the bounding_boxes dict. Extra keys such as 'confidence', 'num_classes', or custom metadata raise a KeyError before shape computation.

Source

Thrown at keras/src/layers/preprocessing/image_preprocessing/max_num_bounding_box.py:98

        boxes = ops.numpy.reshape(boxes, [batch_size, self.max_number, 4])
        labels = ops.numpy.reshape(labels, [batch_size, self.max_number])

        bounding_boxes = bounding_boxes.copy()
        bounding_boxes["boxes"] = boxes
        bounding_boxes["labels"] = labels
        return bounding_boxes

    def transform_segmentation_masks(
        self, segmentation_masks, transformation=None, training=True
    ):
        return self.transform_images(segmentation_masks)

    def compute_output_shape(self, input_shape):
        if isinstance(input_shape, dict) and "bounding_boxes" in input_shape:
            input_keys = set(input_shape["bounding_boxes"].keys())
            extra_keys = input_keys - set(("boxes", "labels"))
            if extra_keys:
                raise KeyError(
                    "There are unsupported keys in `bounding_boxes`: "
                    f"{list(extra_keys)}. "
                    "Only `boxes` and `labels` are supported."
                )

            boxes_shape = list(input_shape["bounding_boxes"]["boxes"])
            boxes_shape[1] = self.max_number
            labels_shape = list(input_shape["bounding_boxes"]["labels"])
            labels_shape[1] = self.max_number
            input_shape["bounding_boxes"]["boxes"] = boxes_shape
            input_shape["bounding_boxes"]["labels"] = labels_shape
        return input_shape

    def get_config(self):
        config = super().get_config()
        config.update({"max_number": self.max_number})
        return config

View on GitHub (pinned to 7a34a03db6)

Solutions

1. Strip extra keys before feeding the layer: bbs = {'boxes': bbs['boxes'], 'labels': bbs['labels']}
2. Carry auxiliary metadata (scores, difficulty flags) outside the bounding_boxes dict, in a parallel structure
3. If you need confidences, fold them into labels or a separate input

Example fix

# before
bbs = {'boxes': boxes, 'labels': labels, 'confidence': confs}
# after
bbs = {'boxes': boxes, 'labels': labels}
confs_outside = confs

allowed = {'boxes', 'labels'}
extra = set(bbs) - allowed
if extra:
    bbs = {k: bbs[k] for k in allowed}
# keep extras separately
extras = {k: v for k, v in bbs_orig.items() if k not in allowed}

def has_only_supported_keys(bbs):
    return set(bbs) <= {'boxes', 'labels'}

When it happens

Trigger: Building a model whose input spec includes {'bounding_boxes': {'boxes': ..., 'labels': ..., 'confidence': ...}} and passing it to this layer's compute_output_shape / model build.

Common situations: Detection pipelines that carry extra per-box metadata alongside labels; converting a COCO-style annotation dict directly to layer input without pruning keys; version changes that tightened accepted keys.

Understand the failure class

Background: Schema validation failed / invalid input schema: payload rejected because its shape doesn't match the expected schema — this error's family across 28 libraries.

Related errors

• `adapt()` can only be called on a tf.data.Dataset or a dict
• Invalid value for argument `output_mode`. Expected one of {a
• `sparse` may only be true if `output_mode` is `"one_hot"`, `
• The `salt` argument for `Hashing` can only be a tuple of siz
• `height` and `width` must be set if `format='xyxy'`.

AI-assisted analysis of keras-team/keras@7a34a03db6 (2026-08-25). Data as JSON: /api/errors/90ec55deff41cd85. Report an issue: GitHub.