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

Unknown iou type {}

Error message

Unknown iou type {}

What it means

COCO results preparation in coco_eval.py dispatches on the iou_type of the COCO evaluator (bbox/segm/keypoints). If the evaluator's iou_type is any other string, prepare() raises ValueError. Detection-only RetinaNet training normally uses 'bbox', so this indicates a misconfigured evaluator.

Source

Thrown at pytorch_object_detection/retinaNet/train_utils/coco_eval.py:66

    def accumulate(self):
        for coco_eval in self.coco_eval.values():
            coco_eval.accumulate()

    def summarize(self):
        for iou_type, coco_eval in self.coco_eval.items():
            print("IoU metric: {}".format(iou_type))
            coco_eval.summarize()

    def prepare(self, predictions, iou_type):
        if iou_type == "bbox":
            return self.prepare_for_coco_detection(predictions)
        elif iou_type == "segm":
            return self.prepare_for_coco_segmentation(predictions)
        elif iou_type == "keypoints":
            return self.prepare_for_coco_keypoint(predictions)
        else:
            raise ValueError("Unknown iou type {}".format(iou_type))

    def prepare_for_coco_detection(self, predictions):
        coco_results = []
        for original_id, prediction in predictions.items():
            if len(prediction) == 0:
                continue

            boxes = prediction["boxes"]
            boxes = convert_to_xywh(boxes).tolist()
            scores = prediction["scores"].tolist()
            labels = prediction["labels"].tolist()

            coco_results.extend(
                [
                    {
                        "image_id": original_id,
                        "category_id": labels[k],
                        "bbox": box,

View on GitHub (pinned to 1ec3fe6f37)

Solutions

1. Check the iouType on the base COCO evaluator before wrapping it in CocoEvaluator; use 'bbox' for detection
2. Pass the base coco_evaluator only after it was created with COCO(..., ) and evaluate(imgs) with iouType='bbox'
3. Add a guard that only builds CocoEvaluator when iou_type in ('bbox','segm','keypoints')

Example fix

// before
coco_evaluator = COCO(...)
coco_evaluator.params.iouType = 'stuff'
// after
coco_evaluator.params.iouType = 'bbox'
coco_evaluator = CocoEvaluator(coco_evaluator, iou_types=['bbox'])

iou_type = coco_evaluator.coco_eval['bbox'].params.iouType if 'bbox' in coco_evaluator.coco_eval else 'unknown'
assert iou_type in ('bbox', 'segm', 'keypoints'), f'unsupported iouType: {iou_type}'

def has_supported_iou_type(evaluator) -> bool:
    return getattr(getattr(evaluator, 'params', None), 'iouType', None) in ('bbox', 'segm', 'keypoints')

try:
    coco_evaluator.update(predictions)
except ValueError as e:
    print(f'iouType misconfigured: {e}; defaulting to bbox')

When it happens

Trigger: Constructing a CocoEvaluator whose base coco_evaluator has iouType set to something other than 'bbox', 'segm', or 'keypoints', then calling update() which invokes prepare().

Common situations: Copying evaluation code for instance segmentation and setting iouType='stuff' or leaving a typo in iouType; mixing evaluation classes from other repos.

Related errors

• not support iou_type: {self.iou_type}
• Unknown iou type {}
• illegal stride value.
• The inverted_residual_setting should not be empty.
• expected stages_repeats as list of 3 positive ints

AI-assisted analysis of WZMIAOMIAO/deep-learning-for-image-processing@1ec3fe6f37 (2026-08-30). Data as JSON: /api/errors/da78b84086ba486a. Report an issue: GitHub.