Description #

Video face blurring operator

Main features #

• Automatically detects faces in video frames and applies blurring
• Supports multiple blur types (mean, box, Gaussian)
• Supports local paths, remote URL/TOS paths, and binary/BASE64 video input
• Consistently outputs the path of the blurred video (the video will be re-encoded and output even if no face is detected)
• Can run in both CPU and GPU environments

Applicable scenarios #

• Privacy protection: De-identifies face information in videos to prevent personal privacy leakage
• Privacy cleansing before data release: Applies blurring to faces in video data before release to protect user privacy
• Face blurring preprocessing for training/evaluation datasets: Applies face blurring to video frames in datasets during model training or evaluation to improve model generalization

Caution #

• Face detection is based on the InsightFace model and is sensitive to lighting and pose variations
• Input videos must be in common formats supported by PyAV/FFmpeg (such as mp4, mov, avi, and more)
• If processing of a single sample fails, the result for that sample returns an empty string and does not affect other samples in the batch

Operator parameters #

Input #

Output #

The output is a column of strings, each representing the file path of the blurred video.

Parameters #

If a parameter does not have a default value, it is required

Examples #

The following code demonstrates how to use daft to run the operator for face blurring in videos.

from __future__ import annotations

import os

import daft
from daft import col
from daft.las.functions.udf import las_udf
from daft.las.functions.video import VideoFaceBlur

if __name__ == "__main__":
    # The modified video will be saved to the specified TOS path. Therefore, you need to set environment variables to ensure write permissions to TOS, including: ACCESS_KEY, SECRET_KEY, TOS_ENDPOINT, TOS_REGION, TOS_TEST_DIR
    TOS_DIR = os.getenv("TOS_DIR", "tos_bucket")
    output_dir = f"tos://{TOS_DIR}/video/video_face_blur"

    if os.getenv("DAFT_RUNNER", "native") == "ray":
        import logging

        import ray

        def configure_logging():
            logging.basicConfig(
                level=logging.INFO,
                format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
                datefmt="%Y-%m-%d %H:%M:%S.%s".format(),
            )
            logging.getLogger("tracing.span").setLevel(logging.WARNING)
            logging.getLogger("daft_io.stats").setLevel(logging.WARNING)
            logging.getLogger("DaftStatisticsManager").setLevel(logging.WARNING)
            logging.getLogger("DaftFlotillaScheduler").setLevel(logging.WARNING)
            logging.getLogger("DaftFlotillaDispatcher").setLevel(logging.WARNING)

        ray.init(dashboard_host="0.0.0.0", runtime_env={"worker_process_setup_hook": configure_logging})
        daft.set_runner_ray()

    daft.set_execution_config(actor_udf_ready_timeout=600)
    daft.set_execution_config(min_cpu_per_task=0)

    tos_dir_url = os.getenv("TOS_DIR_URL", "las-cn-beijing-public-online.tos-cn-beijing.volces.com")
    samples = {"video_path": [f"https://{tos_dir_url}/public/shared_video_dataset/singer.mp4"]}
    ds = daft.from_pydict(samples)

    model_path = os.getenv("MODEL_PATH", "/opt/las/models")
    constructor_kwargs = {
        "output_dir": output_dir,
        "model_path": model_path,
        "blur_type": "gaussian",
        "radius": 20.0,
    }

    ds = ds.with_column(
        "results",
        las_udf(
            VideoFaceBlur,
            construct_args=constructor_kwargs,
            num_gpus=1,
            batch_size=1,
            concurrency=1,
        )(col("video_path")),
    )

    ds.show()

    # ╭────────────────────────────────┬────────────────────────────────╮
    # │ video_path                     ┆ results                        │
    # │ ---                            ┆ ---                            │
    # │ String                         ┆ String                         │
    # ╞════════════════════════════════╪════════════════════════════════╡
    # │ https://las-cn-beijing-public… ┆ tos://las-ai-qa-online/qa/tes… │
    # ╰────────────────────────────────┴────────────────────────────────╯