Struct BatchProcessor 

pub struct BatchProcessor<'a, F>
where
    F: FnMut(Vec<Vec<Results>>, &[DynamicImage], &[String], &[SourceMeta]),
{ /* private fields */ }

A processor for handling batch inference.

This struct manages collecting images into batches, running inference (with fallback), and invoking a callback with the results.

Example

use ultralytics_inference::{YOLOModel, batch::BatchProcessor};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut model = YOLOModel::load("yolo26n.onnx")?;
    let batch_size = 4;
     
    let mut processor = BatchProcessor::new(&mut model, batch_size, |results, images, paths, metas| {
        println!("Processed batch of {} images", results.len());
    });
     
    // Add images...
    // processor.add(image, path, meta);
     
    processor.flush();
    Ok(())
}

Implementations

impl<'a, F> BatchProcessor<'a, F>

where F: FnMut(Vec<Vec<Results>>, &[DynamicImage], &[String], &[SourceMeta]),

pub fn new(model: &'a mut YOLOModel, batch_size: usize, callback: F) -> Self

Create a new BatchProcessor.

Arguments

• model - Mutable reference to the YOLOModel for inference.
• batch_size - Maximum number of images to collect before processing.
• callback - Closure invoked with batch results. Receives:
  • Vec<Vec<Results>> - Results for each image in the batch
  • &[DynamicImage] - The batch images
  • &[String] - Paths for each image
  • &[SourceMeta] - Metadata for each image

Returns

A new BatchProcessor instance.

pub fn add(&mut self, image: DynamicImage, path: String, meta: SourceMeta)

Add an image to the batch.

If the batch becomes full (reaches batch_size), it is automatically processed and the callback is invoked.

Arguments

• image - The image to add.
• path - Path or identifier for this image.
• meta - Source metadata for this image.

pub fn flush(&mut self)

Process any remaining images in the batch.

This should be called after all images have been added to ensure the last partial batch is processed. Has no effect if the batch is empty.