Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in OpenCV BGR format. prompts: Prom

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provides functional-style calling capability.

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style calling capability. Args: img (np.ndarray): Input image in BGR format. Returns:

Provide functional-style calling capability. Args: img (np.ndarray): Input image in BGR format. Returns:

Provide functional-style calling capability. Args: img (np.ndarray): Input image in BGR format. Returns:

Provide functional-style calling capability. Args: img (np.ndarray): Input image in BGR format. image_format

Provide functional-style calling capability. Args: img (np.ndarray): Input image in BGR format. Returns:

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide function-style inference for one image.

Provide function-style inference for one image.

Provide function-style inference for one image.

Provide function-style inference for one image.

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style access to `predict()`. Args: image: Input image in BGR format. resize_type: Optiona

Provide functional-style calling capability, equivalent to ``predict()``. Args: image: Input image in BGR format. Return

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the model and resolve runtime metadata. Args: config (LPRNetConfig): Runtime configuration containing the mod

Initialize the model, load metadata, and precompute grids and anchors. Args: config (YOLOv5Config): Configuration object

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize both CLIP encoder runtimes. Args: config: Runtime configuration containing image and text model paths.

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the model and extract runtime metadata. Args: config: Configuration object containing model path and params.

Initialize both detection and recognition models. Args: config (PaddleOCRConfig): Configuration object containing model paths and

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the FCOS runtime wrapper. Args: config (FCOSConfig): Runtime configuration for FCOS inference. The i

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime and load offline vocabulary embeddings. Args: config: Runtime configuration containing model and

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initializes the model, loads the HBM, and extracts metadata. Args: config (ConvNeXtConfig): Configuration object contain

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the model and extract metadata. Args: config (YOLO26ClsConfig): Configuration object containing model path an

Initialize the model, load metadata, and precompute decoding grids. Args: config (YOLO26PoseConfig): Configuration objec

Initialize the model, load metadata, and precompute decoding grids. Args: config (YOLO26Config): Configuration object co

Initialize the model and extract runtime metadata. Args: config (YOLO26SegConfig): Configuration object containing model

Initialize the model, load metadata, and precompute decoding grids. Args: config (YOLO26OBBConfig): Configuration object

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the classification runtime wrapper. Args: config: Classification runtime configuration for the current model.

Initialize the runtime wrapper and decoding metadata. Args: config: Detection runtime configuration for the current model.

Initialize the segmentation runtime wrapper. Args: config: Segmentation runtime configuration for the current model.

Initialize the pose runtime wrapper. Args: config: Pose runtime configuration for the current model.

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the runtime wrapper and parse model metadata. Args: config: Runtime configuration including model path and in

Initialize the model and extract runtime metadata. Args: config: Configuration object containing model path and params.

Convert a BGR image to NV12 format (Y and UV planes). This function converts a BGR image into NV12 format by first transforming it into plana

Modified forward method for YOLO26 Classify Head (Fully Convolutional).

Modified forward method for YOLO26 Detect Head (BPU-Optimized). Args: x (List[torch.Tensor]): Input feature maps from the neck. Retu

Modified forward method for YOLO26 OBB Head (BPU-Optimized). Args: x (List[torch.Tensor]): Input feature maps from the neck. Returns

Modified forward method for YOLO26 Pose Head (BPU-Optimized). Returns: List[torch.Tensor]: 9 tensors ( [Cls, Box, Kpts] * 3 scales ).

Modified forward method for YOLO26 Segment Head (BPU-Optimized). Returns: List[torch.Tensor]: 10 tensors ( [Cls, Box, MC] * 3 + Proto ).

Crop and rotate a region from an image using a rotated bounding box. This function extracts a rotated rectangular region defined by a four-po

Decode bounding boxes from distributional predictions. This function decodes bounding box coordinates from distribution-based regression outp

Decode keypoint coordinates from model output. This function decodes keypoint predictions from the model output tensor into pixel coordinates

Decode anchor-relative LTRB regression outputs to `xyxy` boxes. This helper converts YOLO-style left, top, right, and bottom distances predic

Decode instance segmentation masks. This function generates instance-level binary masks by linearly combining mask prototype features with pe

Decode all feature maps from the model output. This function iterates over all detection heads, reshapes and reorders the raw output tensors,

Dequantize a dictionary of quantized model outputs. This function applies tensor dequantization to each model output using its corresponding

Ensure that the model file exists locally, downloading it if necessary. If the model file specified by `model_path` does not exist, this function

Draw Top-K results on image.

Log YOLO26 classification results (returns image unchanged).

Draw YOLO26 detection results from `(boxes, scores, cls_ids)` arrays.

Overlay semi-transparent instance masks.

Draw YOLO26 OBB results and log details.

Draw YOLO26 pose results and log details.

Draw YOLO26 segmentation results and log details.

Filter classification outputs using a raw confidence threshold. This function selects classification predictions whose maximum logit value ex

Extract MCES features for selected predictions. This function selects MCES feature vectors corresponding to the given valid prediction indice

Filter detection predictions by confidence threshold. This function combines objectness scores and class probabilities to compute final confi

Extract minimum-area bounding boxes from polygon contours. This function computes the minimum-area bounding rectangle for each polygon contou

Get the SoC (System-on-Chip) name of the current device. The SoC name is read from the system file `/sys/class/boardinfo/soc_name`. If the fi

Load class names from a label file. Each line in the file is treated as a single class name. Empty lines are ignored. Args: path

Load an image from a file path using OpenCV. The image is loaded in BGR color format, which is the default format used by OpenCV. Args:

Load ImageNet labels from a file. This function attempts to load labels that might be stored as a Python dictionary string (e.g., "{0: 'label