YSCV is a pure-Rust computer vision and deep learning framework with SIMD-accelerated tensor operations, image processing, model training, and ONNX inference.
Quick start
use *;
// Create a tensor and run a simple operation.
let a = from_vec.unwrap;
let b = a.scale;
assert_eq!;
// Load an ONNX model and run inference.
// let runner = yscv::onnx::OnnxRunner::load("model.onnx").unwrap();
// let output = runner.run(&[("input", &input_tensor)]).unwrap();
Module overview
| Module | Description |
|---|---|
[tensor] |
N-dimensional tensor with SIMD-accelerated math |
[autograd] |
Automatic differentiation compute graph |
[optim] |
Optimizers (SGD, Adam, AdamW, LAMB, RAdam) |
[kernels] |
Low-level SIMD/GPU compute kernels |
[model] |
Model building, training, data loading, losses |
[imgproc] |
Image I/O, filtering, color, morphology, features |
[video] |
Video capture and frame processing |
[detect] |
Object detection utilities |
[recognize] |
Recognition / embedding search |
[track] |
Multi-object tracking (Hungarian, etc.) |
[eval] |
Evaluation metrics (accuracy, F1, confusion matrix) |
[onnx] |
ONNX model loading and inference |
[cli] |
Command-line interface helpers |