Upload images

One call uploads a folder of images, creating the dataset if it doesn’t exist. A rerun skips the files already uploaded.

from pictograph import Client, UploadReport

client = Client()

report: UploadReport = client.images.upload_from_directory(
    dataset="road-signs",
    source_path="./road_signs",
)
print(f"{report.images_uploaded} uploaded, {report.images_skipped} skipped")

pictograph images upload-directory road-signs ./road_signs

Every argument: upload_from_directory. Raising max_workers past its default of 8 can hit the rate limit.

Where each file lands

A filename is unique within its directory: /stop/001.jpg and /yield/001.jpg are two images.

Re-running and progress

A rerun skips images already in the dataset, so it picks up where a failed one stopped; with skip_existing=False, an existing name is a failure instead. progress is called once per file:

def on_progress(done: int, total: int, filename: str | None) -> None:
    print(f"[{done}/{total}] {filename}")


client.images.upload_from_directory(
    dataset="road-signs",
    source_path="./road_signs",
    progress=on_progress,
)

Errors

A file that fails is listed in report.failures, never raised. The call raises FileNotFoundError when source_path isn’t a directory, and NotFoundError when the dataset doesn’t exist and create_if_missing=False.

Other ways in