ayase 0.1.29

Modular media quality metrics toolkit

Ayase

Modular media quality metrics for video, image, and audio datasets.

Work in progress - APIs and module interfaces may change before 1.0.

What It Does

Ayase runs quality assessment modules over a dataset and writes structured per-sample metrics. 327 modules produce 364 metrics across 19 categories (NR-IQA, FR-IQA, NR-VQA, temporal, motion, audio, face, safety, aesthetics, text-video alignment, and more). Modules are independent - pick only what you need.

Full metric catalog: METRICS.md. Pretrained model catalog: MODELS.md.

Install

pip install ayase

Modules that require ML backends (torch, transformers, pyiqa, insightface, etc.) skip gracefully when the backend is not installed. Model weights are downloaded and cached on first use.

CLI

ayase scan ./dataset                                    # all modules
ayase scan ./dataset --modules metadata,basic_quality   # selected modules
ayase modules list                                      # show all 327 modules
ayase modules check                                     # verify which backends are available
ayase filter ./dataset --min-score 70 --output ./good   # filter by quality
ayase stats ./dataset                                   # dataset statistics
ayase tui                                               # terminal UI

Python API

from ayase import AyasePipeline

pipeline = AyasePipeline(modules=["basic", "metadata", "motion"])
results = pipeline.run("./my_dataset")

for path, sample in results.items():
    qm = sample.quality_metrics
    if qm:
        print(f"{sample.path.name}: technical={qm.technical_score} blur={qm.blur_score}")

pipeline.export("report.json")   # also: report.csv, report.html

Configuration

ayase.toml in project root:

[general]
parallel_jobs = 8

[pipeline]
modules = ["metadata", "basic_quality", "motion"]

[output]
default_format = "json"
artifacts_dir = "reports"

Custom Modules

from ayase.models import Sample, QualityMetrics
from ayase.pipeline import PipelineModule
import cv2

class BlurCheck(PipelineModule):
    name = "blur_check"
    description = "Flag blurry frames via Laplacian variance"
    default_config = {"threshold": 100.0}

    def process(self, sample: Sample) -> Sample:
        img = cv2.imread(str(sample.path), cv2.IMREAD_GRAYSCALE)
        if img is None:
            return sample
        score = float(cv2.Laplacian(img, cv2.CV_64F).var())
        if sample.quality_metrics is None:
            sample.quality_metrics = QualityMetrics()
        sample.quality_metrics.blur_score = score
        if score < self.config.get("threshold", 100.0):
            sample.validation_issues.append({"severity": "warning", "message": f"Blurry ({score:.0f})"})
        return sample

Modules auto-register via __init_subclass__. Config is available as self.config.

Development

git clone <repo-url> && cd ayase
pip install -e ".[dev]"
pytest                    # 8000+ tests, ~4 min
pytest tests/ --full      # with ML model loading

License

MIT. Model weights downloaded at runtime carry their own licenses - see MODELS.md.