What causes "targets, pos_matched_idxs, mask_logits cannot be None when training" in WZMIAOMIAO/deep-learning-for-image-processing?

Ensure targets (with 'masks' and 'labels') is passed in training mode

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

targets, pos_matched_idxs, mask_logits cannot be None when t

Error message

targets, pos_matched_idxs, mask_logits cannot be None when training

What it means

When computing the mask loss in training mode, RoIHeads.forward requires targets, pos_matched_idxs, and mask_logits to all be present; any None makes maskrcnn_loss impossible to compute. The guard raises ValueError naming the missing pieces before unpacking gt_masks/gt_labels.

Source

Thrown at pytorch_object_detection/mask_rcnn/network_files/roi_head.py:546

                # during training, only focus on positive boxes
                num_images = len(proposals)
                mask_proposals = []
                pos_matched_idxs = []
                for img_id in range(num_images):
                    pos = torch.where(labels[img_id] > 0)[0]  # 寻找对应gt类别大于0，即正样本
                    mask_proposals.append(proposals[img_id][pos])
                    pos_matched_idxs.append(matched_idxs[img_id][pos])
            else:
                pos_matched_idxs = None

            mask_features = self.mask_roi_pool(features, mask_proposals, image_shapes)
            mask_features = self.mask_head(mask_features)
            mask_logits = self.mask_predictor(mask_features)

            loss_mask = {}
            if self.training:
                if targets is None or pos_matched_idxs is None or mask_logits is None:
                    raise ValueError("targets, pos_matched_idxs, mask_logits cannot be None when training")

                gt_masks = [t["masks"] for t in targets]
                gt_labels = [t["labels"] for t in targets]
                rcnn_loss_mask = maskrcnn_loss(mask_logits, mask_proposals, gt_masks, gt_labels, pos_matched_idxs)
                loss_mask = {"loss_mask": rcnn_loss_mask}
            else:
                labels = [r["labels"] for r in result]
                mask_probs = maskrcnn_inference(mask_logits, labels)
                for mask_prob, r in zip(mask_probs, result):
                    r["masks"] = mask_prob

            losses.update(loss_mask)

        return result, losses

View on GitHub (pinned to 1ec3fe6f37)

Solutions

Ensure targets (with 'masks' and 'labels') is passed in training mode
Propagate pos_matched_idxs from select_training_samples into the mask-head call
Verify mask_logits are produced before loss computation (mask_head/mask_predictor ran)
Run inference through model.eval() if targets are unavailable

Example fix

// before
roi_heads.forward(images, detections, shapes, targets=None, pos_matched_idxs=None)
// after
proposals, matched_idxs, ..., pos_matched_idxs = roi_heads.select_training_samples(proposals, targets)
roi_heads.forward(images, detections, shapes, targets=targets, pos_matched_idxs=pos_matched_idxs)

Defensive patterns

Strategy: validation

Validate before calling

if roi_heads.training:
    assert targets is not None and pos_matched_idxs is not None and mask_logits is not None
    loss_mask = maskrcnn_loss(mask_logits, mask_proposals, [t['masks'] for t in targets], [t['labels'] for t in targets], pos_matched_idxs)

Type guard

def mask_loss_inputs_ok(targets, pos_matched_idxs, mask_logits):
    return None not in (targets, pos_matched_idxs, mask_logits)

Try / catch

try:
    result, losses = roi_heads(images, detections, shapes, targets, matched_idxs)
except ValueError as e:
    if 'mask_logits' in str(e): logger.error('mask head produced no logits; check has_mask() and inputs')
    raise

Prevention

Keep targets with 'masks' and 'labels' keys in training
Thread pos_matched_idxs through your custom loop
Assert head inputs non-None before loss computation

When it happens

Trigger: Invoking RoIHeads.forward with self.training=True but targets=None, pos_matched_idxs=None, or mask_logits=None — typically from a hand-rolled training loop that skips select_training_samples.

Common situations: Custom training pipelines calling roi_heads internals; forgetting to return/propagate pos_matched_idxs from proposal matching; targets dropped by the dataloader.

Understand the failure class

Background: "missing required argument" and "the following required arguments were not provided": what required-argument errors mean and how to fix them — this error's family across 20 libraries.

Related errors

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/feebf46872d105bc. Report an issue: GitHub.