fix(detection): do not crash from_inference on partial tracker_id (#2353)

process_roboflow_result appended tracker_id only for predictions that carried
one, while xyxy/confidence/class_id were appended for every prediction. A
result where some predictions are tracked and others are not produced a
tracker_id array shorter than the boxes, so Detections.from_inference raised
"tracker_id must be a 1D np.ndarray with shape (N,)".

Collect tracker_id for every prediction (None when absent) and build the array
only when all detections carry one, otherwise leave it None. Fully-tracked and
untracked results are unchanged.

---------

Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
This commit is contained in:
Agis Kounelis
2026-06-26 16:58:52 +02:00
committed by GitHub
co-authored by Jirka Borovec claude[bot]
parent 25d98c85f3
commit 27ba0aa92f
5 changed files with 140 additions and 11 deletions
+2
View File
@@ -7,6 +7,8 @@ date_modified: 2026-06-25
### 0.29.1 <small>Jun 23, 2026</small>
- Fixed [#2353](https://github.com/roboflow/supervision/pull/2353): `sv.Detections.from_inference` no longer raises `TypeError` when the Inference package returns a mixed batch where only some predictions carry a `tracker_id`. `detections.tracker_id` is `None` for the full result in that case; fully-tracked and fully-untracked batches are unchanged.
- Added [#2275](https://github.com/roboflow/supervision/pull/2275): `show_progress: bool = False` parameter to all `sv.DetectionDataset` load and save methods — `from_coco`, `from_yolo`, `from_pascal_voc`, `as_coco`, `as_yolo`, `as_pascal_voc`, and `save_dataset_images`. When `True`, a `tqdm.auto` progress bar is shown (works in terminal and Jupyter). Defaults to `False` for full backward compatibility; no new dependencies.
- Added [#2027](https://github.com/roboflow/supervision/issues/2027): [`sv.InferenceSlicer`](https://supervision.roboflow.com/latest/detection/tools/inference_slicer/#supervision.detection.tools.inference_slicer.InferenceSlicer) now accepts an open rasterio-style dataset in addition to in-memory images. Each tile is read lazily via a windowed read instead of loading the whole image, enabling tiled inference on multi-GB aerial/drone GeoTIFFs without running out of memory. Detection is duck-typed, so `rasterio` stays an optional dependency installable via `pip install "supervision[geotiff]"` and the core library imports no rasterio symbols. A geographic (non-projected) CRS raises `ValueError`.
+7 -1
View File
@@ -635,7 +635,13 @@ class Detections:
`detections.data["class_name"]` is always present as a
string-dtype NumPy array aligned with the detections; it is
empty (shape `(0,)`, dtype str) when `predictions` is empty
or absent.
or absent. `detections.tracker_id` is `None` when no
predictions carry a tracker ID, or when only a subset do
(mixed batch) — in that case all tracker IDs are dropped to
preserve alignment with the bounding boxes. Note: mixed
batches containing both RLE/polygon predictions and box-only
predictions may misalign the `mask` array; this is a
pre-existing limitation not addressed by this fix.
Example:
```python
+17 -10
View File
@@ -74,8 +74,11 @@ def process_roboflow_result(
Returns:
A 6-tuple of ``(xyxy, confidence, class_id, masks, tracker_ids, data)``
where each array is aligned with the others. ``masks`` and
``tracker_ids`` are ``None`` when absent from the predictions.
where each array is aligned with the others. ``masks`` is ``None``
when no predictions include mask data. ``tracker_ids`` is ``None``
when no predictions carry a tracker ID, or when only a subset do
(mixed batch) — in that case all tracker IDs are dropped to preserve
alignment with ``xyxy``.
Examples:
>>> from supervision.detection.utils.internal import process_roboflow_result
@@ -99,7 +102,7 @@ def process_roboflow_result(
class_id: list[int] = []
class_name: list[str] = []
masks: list[npt.NDArray[np.bool_]] = []
tracker_ids: list[int] = []
tracker_ids: list[int | None] = []
image_width = int(roboflow_result["image"]["width"])
image_height = int(roboflow_result["image"]["height"])
@@ -143,15 +146,13 @@ def process_roboflow_result(
class_name.append(prediction["class"])
confidence.append(prediction["confidence"])
masks.append(mask)
if "tracker_id" in prediction:
tracker_ids.append(prediction["tracker_id"])
tracker_ids.append(prediction.get("tracker_id"))
elif "points" not in prediction:
xyxy.append([x_min, y_min, x_max, y_max])
class_id.append(prediction["class_id"])
class_name.append(prediction["class"])
confidence.append(prediction["confidence"])
if "tracker_id" in prediction:
tracker_ids.append(prediction["tracker_id"])
tracker_ids.append(prediction.get("tracker_id"))
elif len(prediction["points"]) >= 3:
polygon = np.array(
[[point["x"], point["y"]] for point in prediction["points"]], dtype=int
@@ -164,8 +165,7 @@ def process_roboflow_result(
class_name.append(prediction["class"])
confidence.append(prediction["confidence"])
masks.append(mask)
if "tracker_id" in prediction:
tracker_ids.append(prediction["tracker_id"])
tracker_ids.append(prediction.get("tracker_id"))
xyxy_arr: npt.NDArray[np.floating] = (
np.array(xyxy, dtype=np.float64) if len(xyxy) > 0 else np.empty((0, 4))
@@ -184,8 +184,15 @@ def process_roboflow_result(
masks_arr: npt.NDArray[np.bool_] | None = (
np.array(masks, dtype=bool) if len(masks) > 0 else None
)
if tracker_ids and 0 < tracker_ids.count(None) < len(tracker_ids):
logger.warning(
"Partial tracker_id in batch; dropping all tracker_ids to preserve "
"alignment with xyxy."
)
tracker_id_arr: npt.NDArray[np.integer] | None = (
np.array(tracker_ids, dtype=np.int64) if len(tracker_ids) > 0 else None
np.array(tracker_ids, dtype=np.int64)
if tracker_ids and None not in tracker_ids
else None
)
data: dict[str, npt.NDArray[np.generic]] = {CLASS_NAME_DATA_FIELD: class_name_arr}
+40
View File
@@ -943,6 +943,46 @@ def test_is_empty(detections: Detections, expected: bool) -> None:
assert detections.is_empty() == expected
def test_from_inference_partial_tracker_id_does_not_crash() -> None:
"""Results where only some predictions carry a tracker_id must not raise."""
result = {
"image": {"width": 200, "height": 200},
"predictions": [
{
"x": 50,
"y": 50,
"width": 20,
"height": 20,
"confidence": 0.9,
"class": "a",
"class_id": 0,
"tracker_id": 7,
},
{
"x": 120,
"y": 120,
"width": 20,
"height": 20,
"confidence": 0.8,
"class": "b",
"class_id": 1,
},
],
}
detections = Detections.from_inference(result)
# all detections are kept; tracker_id is dropped rather than misaligned
assert len(detections) == 2
assert detections.tracker_id is None
assert detections.class_id is not None
assert np.array_equal(detections.class_id, np.array([0, 1]))
assert detections.confidence is not None
assert np.array_equal(detections.confidence, np.array([0.9, 0.8]))
assert detections.xyxy.shape == (2, 4)
assert detections["class_name"] is not None
def test_from_inference_empty_class_name_dtype_matches_non_empty() -> None:
"""Empty and non-empty results should produce string-kind class_name arrays."""
empty_result = {"predictions": [], "image": {"width": 100, "height": 100}}
+74
View File
@@ -449,6 +449,80 @@ TEST_RLE_NONCONTIGUOUS_MASK[0, 3, 2:4] = True
),
DoesNotRaise(),
), # mixed RLE + box-only batch — masks misaligned with xyxy (known limitation)
pytest.param(
{
"predictions": [
{
"x": 1.5,
"y": 1.5,
"width": 2.0,
"height": 2.0,
"confidence": 0.9,
"class_id": 0,
"class": "person",
"tracker_id": 7,
},
{
"x": 3.0,
"y": 3.0,
"width": 2.0,
"height": 2.0,
"confidence": 0.8,
"class_id": 1,
"class": "car",
},
],
"image": {"width": 4, "height": 4},
},
(
np.array([[0.5, 0.5, 2.5, 2.5], [2.0, 2.0, 4.0, 4.0]]),
np.array([0.9, 0.8]),
np.array([0, 1]),
None,
# tracker_id is None when only some detections carry one, rather
# than an array misaligned with xyxy that would raise ValueError.
None,
{CLASS_NAME_DATA_FIELD: np.array(["person", "car"])},
),
DoesNotRaise(),
id="mixed_tracker_id_batch_drops_to_none",
),
pytest.param(
{
"predictions": [
{
"x": 1.5,
"y": 1.5,
"width": 2.0,
"height": 2.0,
"confidence": 0.9,
"class_id": 0,
"class": "person",
},
{
"x": 3.0,
"y": 3.0,
"width": 2.0,
"height": 2.0,
"confidence": 0.8,
"class_id": 1,
"class": "car",
},
],
"image": {"width": 4, "height": 4},
},
(
np.array([[0.5, 0.5, 2.5, 2.5], [2.0, 2.0, 4.0, 4.0]]),
np.array([0.9, 0.8]),
np.array([0, 1]),
None,
# None not in tracker_ids prevents np.array([None, None], dtype=int64)
None,
{CLASS_NAME_DATA_FIELD: np.array(["person", "car"])},
),
DoesNotRaise(),
id="all_absent_tracker_id_no_raise",
),
],
)
def test_process_roboflow_result(