CLI & Batch

ofiqpy — single / small runs

Mirrors OFIQSampleApp's interface:

ofiqpy -i <image|directory> -o <out.csv>

• -i accepts a single image file or a directory (searched recursively for .jpg/.jpeg/.png/.bmp).
• -o writes an OFIQ-format CSV.

python -m ofiqpy.batch — parallel batch

For large directories, the batch runner processes images across worker processes and is resumable:

python -m ofiqpy.batch -i <directory> -o <out.csv> [-w N] [--resume]

Flag	Meaning
-i	input directory (recursive) or file
-o	output CSV
-w	worker processes (default: number of CPUs)
--resume	skip images already present in the output CSV

Each worker builds its own pipeline (ONNX / cv2.ml models are not shared across processes). Rows are flushed as they complete, so an interrupted run can be resumed with --resume.

Output format

Semicolon-delimited, matching OFIQSampleApp:

Filename;UnifiedQualityScore;BackgroundUniformity;...;<component>;...;
        UnifiedQualityScore.scalar;...;<component>.scalar;...;assessment_time_in_ms

• One raw (native) column and one .scalar (0–100) column per component, in OFIQ's order.
• A component that fails to assess emits the OFIQ sentinel: raw 0, scalar -1.
• Load with pandas:

import pandas as pd
df = pd.read_csv("out.csv", sep=";")
df["UnifiedQualityScore.scalar"].describe()