Skip to content

infer_batch

Run inference over an image directory and persist the detections as a reusable manifest.

Inputs

  • checkpoint
  • input_dir
  • output_dir
  • imgsz, conf, device, batch_size
  • save_annotated
  • export_format — json, csv, or yolo_labels

Outputs

  • number of processed images
  • total detection count
  • manifest path for the chosen export format
  • optional annotated image directory
  • a small preview list for Studio or CLI summaries

Examples

{"checkpoint":"yolov8n.pt","input_dir":"~/images"}
{"checkpoint":"run_demo","input_dir":"~/images","export_format":"csv","save_annotated":false}

Common Errors

  • input directory missing
  • no supported images found under the input directory

infer_image, benchmark_latency, model_profile