open-mmlab/mmdetection · error · NotImplementedError

Unknown anchor type: {self.anchor_type}.Please use `anchor_f

Error message

Unknown anchor type: {self.anchor_type}.Please use `anchor_free` or `anchor_based`.

What it means

Error "Unknown anchor type: {self.anchor_type}.Please use `anchor_free` or `anchor_based`." thrown in open-mmlab/mmdetection.

Source

Thrown at mmdet/models/dense_heads/tood_head.py:284

            cls_logits = self.tood_cls(cls_feat)
            cls_prob = self.cls_prob_module(feat)
            cls_score = sigmoid_geometric_mean(cls_logits, cls_prob)

            # reg prediction and alignment
            if self.anchor_type == 'anchor_free':
                reg_dist = scale(self.tood_reg(reg_feat).exp()).float()
                reg_dist = reg_dist.permute(0, 2, 3, 1).reshape(-1, 4)
                reg_bbox = distance2bbox(
                    self.anchor_center(anchor) / stride[0],
                    reg_dist).reshape(b, h, w, 4).permute(0, 3, 1,
                                                          2)  # (b, c, h, w)
            elif self.anchor_type == 'anchor_based':
                reg_dist = scale(self.tood_reg(reg_feat)).float()
                reg_dist = reg_dist.permute(0, 2, 3, 1).reshape(-1, 4)
                reg_bbox = self.bbox_coder.decode(anchor, reg_dist).reshape(
                    b, h, w, 4).permute(0, 3, 1, 2) / stride[0]
            else:
                raise NotImplementedError(
                    f'Unknown anchor type: {self.anchor_type}.'
                    f'Please use `anchor_free` or `anchor_based`.')
            reg_offset = self.reg_offset_module(feat)
            bbox_pred = self.deform_sampling(reg_bbox.contiguous(),
                                             reg_offset.contiguous())

            # After deform_sampling, some boxes will become invalid (The
            # left-top point is at the right or bottom of the right-bottom
            # point), which will make the GIoULoss negative.
            invalid_bbox_idx = (bbox_pred[:, [0]] > bbox_pred[:, [2]]) | \
                               (bbox_pred[:, [1]] > bbox_pred[:, [3]])
            invalid_bbox_idx = invalid_bbox_idx.expand_as(bbox_pred)
            bbox_pred = torch.where(invalid_bbox_idx, reg_bbox, bbox_pred)

            cls_scores.append(cls_score)
            bbox_preds.append(bbox_pred)
        return tuple(cls_scores), tuple(bbox_preds)

View on GitHub (pinned to cfd5d3a985)

When it happens

Trigger: Thrown at mmdet/models/dense_heads/tood_head.py:284 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/e9eebf6be0c1653b. Report an issue: GitHub.