WZMIAOMIAO/deep-learning-for-image-processing · error · TypeError

mask_roi_pool should be of type MultiScaleRoIAlign or None i

Error message

mask_roi_pool should be of type MultiScaleRoIAlign or None instead of {}

What it means

MaskRCNN.__init__ type-checks mask_roi_pool: it must be a MultiScaleRoIAlign instance or None (to use the default). Any other object raises this TypeError reporting the actual type received.

Source

Thrown at pytorch_object_detection/mask_rcnn/network_files/mask_rcnn.py:138

            box_roi_pool=None,
            box_head=None,
            box_predictor=None,
            box_score_thresh=0.05,
            box_nms_thresh=0.5,
            box_detections_per_img=100,
            box_fg_iou_thresh=0.5,
            box_bg_iou_thresh=0.5,
            box_batch_size_per_image=512,
            box_positive_fraction=0.25,
            bbox_reg_weights=None,
            # Mask parameters
            mask_roi_pool=None,
            mask_head=None,
            mask_predictor=None,
    ):

        if not isinstance(mask_roi_pool, (MultiScaleRoIAlign, type(None))):
            raise TypeError(
                f"mask_roi_pool should be of type MultiScaleRoIAlign or None instead of {type(mask_roi_pool)}"
            )

        if num_classes is not None:
            if mask_predictor is not None:
                raise ValueError("num_classes should be None when mask_predictor is specified")

        out_channels = backbone.out_channels

        if mask_roi_pool is None:
            mask_roi_pool = MultiScaleRoIAlign(featmap_names=["0", "1", "2", "3"], output_size=14, sampling_ratio=2)

        if mask_head is None:
            mask_layers = (256, 256, 256, 256)
            mask_dilation = 1
            mask_head = MaskRCNNHeads(out_channels, mask_layers, mask_dilation)

        if mask_predictor is None:

View on GitHub (pinned to 1ec3fe6f37)

Solutions

1. Use torchvision.ops.MultiScaleRoIAlign(featmap_names=['0','1','2','3'], output_size=14, sampling_ratio=2) for mask_roi_pool
2. Pass None to let the model build the default MultiScaleRoIAlign
3. If you need custom pooling, subclass or wrap it so the argument is a genuine MultiScaleRoIAlign, or patch the check consciously

Example fix

// before
from torchvision.ops import RoIAlign
model = MaskRCNN(backbone, num_classes=91, mask_roi_pool=RoIAlign((14,14), 2, 0))
// after
from torchvision.ops import MultiScaleRoIAlign
model = MaskRCNN(backbone, num_classes=91,
                 mask_roi_pool=MultiScaleRoIAlign(featmap_names=['0','1','2','3'], output_size=14, sampling_ratio=2))

if mask_roi_pool is not None and not isinstance(mask_roi_pool, MultiScaleRoIAlign):
    raise TypeError(f'mask_roi_pool must be MultiScaleRoIAlign or None, got {type(mask_roi_pool)}')

def is_valid_mask_roi_pool(x):
    return x is None or isinstance(x, MultiScaleRoIAlign)

try:
    model = MaskRCNN(backbone, num_classes=num_classes, mask_roi_pool=mask_roi_pool)
except TypeError as e:
    if 'mask_roi_pool' in str(e):
        model = MaskRCNN(backbone, num_classes=num_classes, mask_roi_pool=None)
    else:
        raise

When it happens

Trigger: Passing a custom ROI-pooling module (e.g. RoIAlign from torchvision.ops, an nn.AdaptiveAvgPool2d, or a differently-parameterized pooling layer) as the mask_roi_pool argument of MaskRCNN(...).

Common situations: Confusing torchvision.ops.RoIAlign with MultiScaleRoIAlign; copying a box_roi_pool instance into mask_roi_pool from the wrong class hierarchy; older/newer torchvision versions where the expected class moved modules.

Understand the failure class

Background: "Wrong argument type", "must be a string", "expected Array or Prism::Scope": TypeError and ArgumentError when a library receives a value of the wrong type — this error's family across 28 libraries.

Related errors

• In training mode, targets should be passed
• Expected target boxes to be of type Tensor, got {:}.
• num_classes should be None when box_predictor is specified
• num_classes should not be None when box_predictor is not spe
• In training mode, targets should be passed

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/1a40f518d8d5887f. Report an issue: GitHub.