Description #

Image quality scoring processor, evaluates image visual quality based on CLIP-IQA

Key features: #

• Quality scoring: Outputs a score in the range of 0–1, with higher scores indicating better visual quality
• Multi-source support: Supports input via URL, TOS, local path, binary stream, and more
• Stable processing: Processes each image individually, suitable for multi-source image input
• Configurable prompts: Supports custom prompts

Format support: #

• Input: Supports multiple formats such as image URL, TOS address, binary stream
• Output: Floating-point quality score (0.0–1.0)
• Image formats: Mainstream image formats such as JPG, PNG, WebP
• Resolution: Automatically adapts to images of different resolutions

Scoring principle: #

• Visual feature extraction: Uses the CLIP model to extract image features
• Prompt comparison: Calculates similarity between the image and the prompt
• Probability output: Returns the probability that the image is closer to the positive prompt

Operator parameters #

Input #

Output #

Floating-point array containing quality scores, score range 0.0–1.0.
Images that cannot be decoded or processed return a null value

Parameters #

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

Examples #

The following code demonstrates how to use Daft (for distributed scenarios) to run the operator for image quality scoring.

from __future__ import annotations

import os

import daft
from daft import col
from daft.las.functions.image import ImageQualityScore
from daft.las.functions.udf import las_udf

if __name__ == "__main__":
    TOS_TEST_DIR_URL = os.getenv("TOS_TEST_DIR_URL", "las-cn-beijing-public-online.tos-cn-beijing.volces.com")
    model_path = os.getenv("MODEL_PATH", "/opt/las/models")

    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",
            )
            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)

    samples = {
        "input_path": [f"https://{TOS_TEST_DIR_URL}/public/shared_image_dataset/forest.jpg"],
    }
    ds = daft.from_pydict(samples)

    constructor_kwargs = {
        "model_path": model_path,
        "clip_model_name": "openai/clip-vit-large-patch14",
        "prompt": "quality",
    }

    ds = ds.with_column(
        "quality_score",
        las_udf(
            ImageQualityScore,
            construct_args=constructor_kwargs,
            num_gpus=1,
            concurrency=1,
            batch_size=1,
        )(col("input_path")),
    )

    ds.show()
    # ╭────────────────────────────────┬────────────────────╮
    # │ input_path                     ┆ quality_score      │
    # │ ---                            ┆ ---                │
    # │ String                         ┆ Float64            │
    # ╞════════════════════════════════╪════════════════════╡
    # │ https://las-cn-beijing-publi-… ┆ 0.39               │
    # ╰────────────────────────────────┴────────────────────╯