obb_loss

Typed PyTorch oriented-box losses and standalone matching for DETR-like detectors. The package exposes pure functions, Pydantic-configured nn.Module criteria, and a typed callback composition. Detector-layer orchestration stays in your pipeline.

See runnable examples for complete CPU forward/backward scripts.

The primary geometry is (cx, cy, width, height, cos(theta), sin(theta)), with positive sides and nonzero direction vectors. Coordinate units are explicit and shared by predictions and targets. Use boxes_from_angles to convert radians; no format inference occurs.

Pack existing tensors

import obb_loss as obb
import torch

# Existing per-sample target tensors, including an empty sample.
targets_result = obb.pack_boxes(
    [torch.tensor([[0., 0., 2., 1., 1., 0.]]), torch.empty(0, 6)],
    device=torch.device("cpu"),
    dtype=torch.float32,
)
# Dense detector output [batch, queries, 6] also has a helper.
predictions = torch.tensor([[[0.1, 0., 2., 1., 1., 0.]],
                            [[0.2, 0., 2., 1., 1., 0.]]], requires_grad=True)
preds_result = obb.pack_dense_boxes(predictions)

Helpers preserve gradients and sample boundaries, never mutate or silently transfer inputs, and return Ok(PackedBoxes) or Err(InputError). Device and dtype remain explicit even for zero-sample input. PackedBoxes.boxes is [total_boxes, 6]; cu_seqlens is integer [batch + 1], starts at zero, ends at total_boxes, and retains empty samples through repeats.

Match, then evaluate an aligned box loss

import obb_loss as obb

match (preds_result, targets_result):
    case (obb.Ok(value=preds), obb.Ok(value=targets)):
        matcher = obb.HungarianMatcher(obb.MatcherConfig(chunk_size=4096, max_pairs=1_000_000))
        match matcher.forward(preds, targets):
            case obb.Ok(value=assignment):
                matched_preds, matched_targets = obb.gather_matched(preds, targets, assignment)
                criterion = obb.ProbIoULoss(obb.ProbIoUConfig(reduction=obb.Reduction.MEAN))
                loss = criterion.forward(matched_preds, matched_targets)
                loss.backward()
            case obb.Err(error=error):
                print(error)
    case _:
        print("Packing failed; inspect the typed error outcome")

Box criterion signature: forward(preds: PackedBoxes, targets: PackedBoxes) -> Tensor. forward trusts geometry, aligned counts, and offsets; checked validates fresh inputs and returns a typed result. Pure stable_probiou_loss returns unreduced values and stable_probiou returns similarity. Outputs are float64; autocast is disabled locally without changing your detector's outer mixed-precision context. Empty mean/sum is a gradient-connected zero. Mean averages matched pairs across the whole pack, not per-sample means.

The matcher solves each sample independently, assigns the smaller side fully, and returns detached indices on the input device. Its CPU solver makes an explicit cost transfer from CUDA. Cost construction is chunked; complete cost storage remains quadratic within each sample. See matching contracts for limits and solver references.

Quality classification

For one complete decoder layer, obb.SetCriterion owns matching and weighted classification, L1, and ProbIoU losses. Use obb.pack_predictions(boxes, logits) and obb.pack_targets(samples, device=..., dtype=...), where each obb.TargetSample contains boxes, class labels, and caller-supplied quality. Match obb.Ok(value=losses) from criterion.forward(preds, targets) and backpropagate losses.total. Pass reduction=obb.Reduction.NONE for [batch] losses, or SUM/MEAN for a scalar. Normalization is explicit: normalizer=obb.TargetCount.PER_SAMPLE, PER_BATCH, or a positive scalar. The defaults SUM/PER_BATCH preserve batch-wide target normalization. Pass named callbacks through SetCriterion(losses=..., matcher=...); use obb.WeightedLoss(callback, obb.WeightConfig(weight=...)) to configure each term. Matcher cost callbacks own their weights independently. Results expose normalized weighted values through losses.terms[name] and their sum through losses.total. Extend obb.default_set_losses() with your own callback, including query regularizers over unmatched predictions. See the custom set loss, the DETR example, and set criterion contracts.

import obb_loss as obb

logits = torch.tensor([[0.3, -0.8]], requires_grad=True)
quality = torch.tensor([[0.7, 0.]])  # Caller-supplied [queries, classes], zero = negative.
vfl = obb.VarifocalLoss(obb.VFLConfig(alpha=0.75, gamma=2., quality_weighted=True))
qfl = obb.QualityFocalLoss(obb.QFLConfig(beta=2.))
classification_loss = vfl.forward(logits, quality)  # Or qfl.forward(logits, quality).

Pure varifocal_loss and quality_focal_loss return unreduced [queries, classes] values. Checked modules reject nonfinite logits and quality outside [0, 1]. Targets are not automatically detached; choose that policy explicitly. Mean averages all class elements; positive-count or distributed normalization is yours. No auxiliary decoder-layer loop or classification target synthesis is included. See loss papers and numerical policy.

Custom functions and typed settings

import obb_loss as obb
from torch import Tensor

def center_loss(context: obb.BoxLossContext) -> obb.Result[Tensor, obb.InputError]:
    """Return one squared center penalty per aligned packed pair."""
    delta = context.preds.boxes[:, :2].double() - context.targets.boxes[:, :2].double()
    return obb.Ok(delta.square().sum(-1))

criterion = obb.CompositeBoxLoss(losses={
    "probiou": obb.box_probiou_loss,
    "center": obb.WeightedLoss(center_loss, obb.WeightConfig(weight=0.1)),
})
# criterion(preds, targets, reduction=obb.Reduction.MEAN) returns Ok(loss) or Err.
config_result = obb.parse_config(obb.ProbIoUConfig, {"weight": 2., "reduction": "sum"})

Box and set compositions share typed Result-returning callbacks and independently weighted terms. Box callbacks receive aligned packed geometry through BoxLossContext; set callbacks receive SetLossContext, including unmatched predictions. Conversion helpers pack list/dense inputs at the caller boundary. Reduction is applied once by the criterion. Callback bugs propagate, while observable output failures return Err. Module parameters, buffers and hooks are preserved; zero-weight terms are omitted. JSON configuration never serializes live callbacks.

The former BoxLossBuilder and build_loss/build_matcher factories are removed. Construct CompositeBoxLoss from a callback mapping and built-in modules directly. Composite box forward now returns Result and takes reduction at evaluation time. Callback module state paths are shorter; recreate composition before loading a compatible state dict.

parse_config accepts JSON-compatible values and returns typed validation errors for external settings. Direct Pydantic model construction retains its exception semantics.

Verification

Use uv sync to install the development environment and just check to run strict formatting, lint, type, and test checks. Python development is pinned to 3.12; package metadata supports Python 3.11+. Tool settings live separately in ruff.toml, pyrefly.toml, and pytest.ini.

Tensor annotations use native Pyrefly shapes. Shape packages are development-only, versioned with the checker, and installed on Python 3.12+. Shape imports stay under TYPE_CHECKING, so Python 3.11 runtime users need neither package. just types uses tools.typecheck to provision an isolated Python 3.12 checker when invoked from 3.11; the checker still validates Python 3.11 syntax. CI retains runtime tests for 3.11–3.14. See typing contracts for guarantees and experimental stub limitations.

just stress --samples 10000 --chunk-size 512 --dtype FP64
just benchmark --samples 4 --queries 128 --targets 32
just test-cuda
just build

Stress and benchmark commands emit JSON on stdout and progress on stderr; use --quiet to suppress progress. test-cuda fails if hardware is unavailable, so skipped cases cannot masquerade as GPU verification. Read verification evidence for measured results and limitations. Build creates a portable pure-Python wheel; nothing publishes.