Crate wifi_densepose_nn 

WiFi-DensePose Neural Network Crate

This crate provides neural network inference capabilities for the WiFi-DensePose pose estimation system. It supports multiple backends including ONNX Runtime, tch-rs (PyTorch), and Candle for flexible deployment.

Features

• DensePose Head: Body part segmentation and UV coordinate regression
• Modality Translator: CSI to visual feature space translation
• Multi-Backend Support: ONNX, PyTorch (tch), and Candle backends
• Inference Optimization: Batching, GPU acceleration, and model caching

Example

use wifi_densepose_nn::{InferenceEngine, DensePoseConfig, OnnxBackend};

// Create inference engine with ONNX backend
let config = DensePoseConfig::default();
let backend = OnnxBackend::from_file("model.onnx")?;
let engine = InferenceEngine::new(backend, config)?;

// Run inference
let input = ndarray::Array4::zeros((1, 256, 64, 64));
let output = engine.infer(&input)?;

Re-exports

pub use densepose::DensePoseConfig;

pub use densepose::DensePoseHead;

pub use densepose::DensePoseOutput;

pub use error::NnError;

pub use error::NnResult;

pub use inference::Backend;

pub use inference::InferenceEngine;

pub use inference::InferenceOptions;

pub use onnx::OnnxBackend;

pub use onnx::OnnxSession;

pub use tensor::Tensor;

pub use tensor::TensorShape;

pub use translator::ModalityTranslator;

pub use translator::TranslatorConfig;

pub use translator::TranslatorOutput;

Modules

densepose

DensePose head for body part segmentation and UV coordinate regression.

error

Error types for the neural network crate.

inference

Inference engine abstraction for neural network backends.

onnx

ONNX Runtime backend for neural network inference.

prelude

Prelude module for convenient imports

tensor

Tensor types and operations for neural network inference.

translator

Modality translation network for CSI to visual feature space conversion.

Constants

DEFAULT_HIDDEN_CHANNELS

Default hidden channel sizes for networks

NUM_BODY_PARTS

Number of body parts in DensePose model (standard configuration)

NUM_UV_COORDINATES

Number of UV coordinates (U and V)

VERSION

Version information