Main entry point for pose model export.

Run COCO evaluation.

Run YOLO26 inference on a single image. This function orchestrates the complete inference process: - Argument parsing - Task wrapper init

Run the complete RepViT classification pipeline on a single image. The entry follows the standardized sample pattern used in this repository

Run Ultralytics YOLO inference on a single image. This function orchestrates the complete inference process: - Argument parsing - Task wr

Run the complete MobileNetV4 classification pipeline on a single image. The entry follows the standardized sample pattern used in this repos

Run the complete GoogLeNet classification pipeline on a single image. The entry follows the standardized sample pattern used in this reposit

Run the complete MODNet matting pipeline on a single image.

Parse comma-separated prompt words. Args: prompts: Comma-separated prompt string. Returns: Prompt list with empty items

Parse comma-separated text prompts from the command line. Args: texts: Comma-separated prompt string. Returns: Prompt l

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Decode the logits tensor into the final plate string. Args: logits (np.ndarray): Raw logits tensor returned by the runti

Convert raw model outputs to final detection results. Args: outputs (Dict[str, Dict[str, np.ndarray]]): Raw model output

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Compute cosine similarity between image and text features. Args: image_feature: Image feature vector from the BPU image

Convert the raw probability tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert raw model outputs to final segmentation results. Args: outputs: Raw model outputs. ori_img_w: Original image

Run DB post-processing: binarize, find contours, dilate, compute boxes. Args: outputs (Dict[str, Dict[str, np.ndarray]]): Raw out

Squeeze raw output to a 2-D segmentation map. Args: raw: Raw BPU output of shape (1, H, W, 1) int32. Returns:

Convert the raw probability tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()

Decode FCOS outputs into final detections. Args: outputs (Dict[str, Dict[str, np.ndarray]]): Raw inference outputs.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert YOLOWorld raw outputs into final detection arrays. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Converts raw BPU output tensors to classification results. Args: outputs (Any): Raw output tensors from forward().

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert classification logits into Top-K prediction results. Args: outputs: Raw runtime outputs returned by `forward`. R

Decode detection tensors into final sample results. Args: outputs: Raw runtime outputs returned by `forward`. ori_img

Decode segmentation tensors into boxes, classes, and masks. Args: outputs: Raw runtime outputs returned by `forward`.

Decode pose tensors into boxes, scores, and keypoints. Args: outputs: Raw runtime outputs returned by `forward`. ori_

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert the raw logits tensor into Top-K classification results. Args: outputs: Raw runtime outputs from `forward()`.

Convert raw model output to an alpha matte in original image space. Args: outputs: Raw model outputs. Returns:

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Load the packed `float32` tensor used as the LPRNet input. Returns: Dict[str, Dict[str, np.ndarray]]: Nested input dicti

Convert a BGR image to packed NV12 input for hbm_runtime. Args: image (np.ndarray): Input image in BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the image encoder tensor. Args: image: Input image in OpenCV BGR format. Returns:

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert a BGR image to packed NV12 input for hbm_runtime. Args: img: Input image in BGR format. resize_type: Override

Convert input image to NV12 for the detection model. Args: image (np.ndarray): Input image in BGR format (H, W, 3). Retu

Convert BGR image to NV12 tensor for BPU input. Args: bgr: Input image as (H, W, 3) uint8 BGR array. Returns:

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert a BGR image to packed NV12 input for `hbm_runtime`. Args: image (np.ndarray): Input BGR image. resiz

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image and prompt list into model inputs. Args: image: Input image in OpenCV BGR format. prom

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Converts input image to the format required by BPU runtime (NV12). Args: image (np.ndarray): Input image in BGR format (

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into packed NV12 model input.

Convert one BGR image into packed NV12 model input. Args: img: Input BGR image loaded by OpenCV. resize_type: Optiona

Convert one BGR image into packed NV12 model input. Args: img: Input BGR image loaded by OpenCV. resize_type: Optiona

Convert one BGR image into packed NV12 model input.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert one BGR image into the packed NV12 tensor expected by the model. Args: image: Input image in OpenCV BGR format.

Convert a BGR image to float32 NCHW RGB input for hbm_runtime. The preprocessing pipeline: 1. BGR -> RGB 2. Normalize to [-1,

End-to-end inference: BGR image -> segmentation map. Args: bgr: Input image as (H, W, 3) uint8 BGR array. Returns:

Run the complete classification pipeline on a single image. This method orchestrates the full workflow: - Pre-process the in

Run the complete pose inference pipeline on a single image. This method orchestrates the full workflow: - Pre-process the in

High-level interface for one-click inference. Args: img (np.ndarray): Input image. Returns: Detecti

Run the complete segmentation pipeline on a single image. Args: img (np.ndarray): Input image. image_format