open-mmlab/mmdetection · error · ValueError

Only support 300 or 512 in SSDAnchorGenerator when not setti

Error message

Only support 300 or 512 in SSDAnchorGenerator when not setting min_sizes and max_sizes, got {self.input_size}.

What it means

When min_sizes and max_sizes are not supplied, SSDAnchorGenerator can only auto-generate anchors for the two canonical SSD input sizes: 300 and 512. Any other input_size has no size preset and the constructor rejects it.

Source

Thrown at mmdet/models/task_modules/prior_generators/anchor_generator.py:577

                    raise ValueError(
                        'basesize_ratio_range[0] should be either 0.15'
                        'or 0.2 when input_size is 300, got '
                        f'{basesize_ratio_range[0]}.')
            elif self.input_size == 512:
                if basesize_ratio_range[0] == 0.1:  # SSD512 COCO
                    min_sizes.insert(0, int(self.input_size * 4 / 100))
                    max_sizes.insert(0, int(self.input_size * 10 / 100))
                elif basesize_ratio_range[0] == 0.15:  # SSD512 VOC
                    min_sizes.insert(0, int(self.input_size * 7 / 100))
                    max_sizes.insert(0, int(self.input_size * 15 / 100))
                else:
                    raise ValueError(
                        'When not setting min_sizes and max_sizes,'
                        'basesize_ratio_range[0] should be either 0.1'
                        'or 0.15 when input_size is 512, got'
                        f' {basesize_ratio_range[0]}.')
            else:
                raise ValueError(
                    'Only support 300 or 512 in SSDAnchorGenerator when '
                    'not setting min_sizes and max_sizes, '
                    f'got {self.input_size}.')

        assert len(min_sizes) == len(max_sizes) == len(strides)

        anchor_ratios = []
        anchor_scales = []
        for k in range(len(self.strides)):
            scales = [1., np.sqrt(max_sizes[k] / min_sizes[k])]
            anchor_ratio = [1.]
            for r in ratios[k]:
                anchor_ratio += [1 / r, r]  # 4 or 6 ratio
            anchor_ratios.append(torch.Tensor(anchor_ratio))
            anchor_scales.append(torch.Tensor(scales))

        self.base_sizes = min_sizes
        self.scales = anchor_scales

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Set input_size to 300 or 512, or
2. Provide explicit min_sizes and max_sizes lists computed for your resolution
3. Check that input_size is a plain int (not a tuple like (512, 512)) which would fail the equality checks

Example fix

# before
anchor_generator=dict(type='SSDAnchorGenerator', input_size=640)
# after
anchor_generator=dict(
    type='SSDAnchorGenerator', input_size=640,
    min_sizes=[[21, 45], [102, 216], [267, 411], [399, 563], [534, 704], [669, 887]],
    max_sizes=[[45, 99], [216, 336], [411, 549], [563, 721], [704, 880], [887, 1055]])

ag = cfg['anchor_generator']
assert isinstance(ag.get('input_size'), int) and (ag['input_size'] in (300, 512) or ag.get('min_sizes'))

When it happens

Trigger: SSDAnchorGenerator with input_size set to anything other than 300 or 512 (e.g. 320, 512x512 as a tuple, 800) while omitting min_sizes/max_sizes.

Common situations: Resizing SSD training to a non-standard resolution, or using a tuple input_size, without hand-specifying anchor sizes.

Related errors

• basesize_ratio_range[0] should be either 0.15or 0.2 when inp
• When not setting min_sizes and max_sizes,basesize_ratio_rang
• center_offset should be in range [0, 1], {center_offset} is
• Either scales or octave_base_scale with scales_per_octave sh
• LoadImageFromFile is not found in the test pipeline

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/14e1dbe8dea59e55. Report an issue: GitHub.