trtutils.models.hand_interaction package

Module contents

Hand-object interaction model implementations.

class trtutils.models.hand_interaction.HOIDETR(engine_path: Path | str, warmup_iterations: int = 10, input_range: tuple[float, float] = (0.0, 1.0), preprocessor: str = 'trt', resize_method: str = 'letterbox', conf_thres: float = 0.3, pair_thres: float = 0.5, second_pair_thres: float | None = None, nms_iou_thres: float = 0.5, mean: tuple[float, float, float] | None = (0.485, 0.456, 0.406), std: tuple[float, float, float] | None = (0.229, 0.224, 0.225), dla_core: int | None = None, device: int | None = None, backend: str = 'auto', *, warmup: bool | None = None, pagelocked_mem: bool | None = None, unified_mem: bool | None = None, cuda_graph: bool | None = None, no_warn: bool | None = None, verbose: bool | None = None)

Bases: HandInteractionDetector, Model

Alias of HandInteractionDetector with default args for HOI-DETR.

class trtutils.models.hand_interaction.Hands23(engine_path: Path | str, warmup_iterations: int = 10, input_range: tuple[float, float] = (0.0, 255.0), preprocessor: str = 'trt', resize_method: str = 'letterbox', conf_thres: float = 0.3, pair_thres: float = 0.3, second_pair_thres: float | None = 0.7, nms_iou_thres: float = 0.5, mean: tuple[float, float, float] | None = None, std: tuple[float, float, float] | None = None, dla_core: int | None = None, device: int | None = None, backend: str = 'auto', *, warmup: bool | None = None, pagelocked_mem: bool | None = None, unified_mem: bool | None = None, cuda_graph: bool | None = False, no_warn: bool | None = None, verbose: bool | None = None)

Bases: HandInteractionDetector, Model

Alias of HandInteractionDetector with default args for Hands23.

Normalization (BGR flip, mean/std) is baked into the exported ONNX, so the wrapper passes raw 0-255 RGB input straight through. A single conf_thres is used for all three classes, unlike the reference demo which uses per-class thresholds of 0.7/0.5/0.3. CUDA graphs default to off: the in-graph NMS yields data-dependent shapes, which TensorRT cannot capture.