Learning to See with Less: A Survey on Computer Vision with Limited and Imperfect Data

TargetPrimaryModerate confidenceJudged

Survey synthesizes methodological advances in computer vision under constrained data including data augmentation, semi-supervised learning, transfer learning and self-supervised techniques

Moderate confidence

The alignment is plausible and defensible but rests on an assumption a reasonable reviewer could decline. Published, and surfaced in the review queue for a human to confirm or override.

Judged

Judged against the published classification rubric, under the rubric and prompt versions recorded with the run that produced it.

Produced by

claude-haiku-4-5-20251001, under kortex-publication-v1

TargetSecondaryHigh confidenceRule-derivedrule sdg17-structural-v1

Structural signal: co-authorship outside India.

High confidence

The subject matter is a strong and direct inference from the record type, the funding or organising body, or the programme context, even where the title alone is thin. Example: an ICMR-funded project against SDG 3, on the strength of the agency.

Rule-derived

Emitted by a deterministic rule from structural signals rather than judged from the record’s text. The rule carries a version, so rule-derived and judged labels stay separable.

Produced by

sdg17-structural-v1, under kortex-publication-v1

ThemePrimaryModerate confidenceJudged

Derived through the seeded SDG-to-Viksit-Bharat crosswalk from the primary SDG 9 label. Survey synthesizes methodological advances in computer vision under constrained data including data augmentation, semi-supervised learning, transfer learning and self-supervised techniques

Produced by

claude-haiku-4-5-20251001, under kortex-publication-v1

Sub-themeSecondaryHigh confidenceJudged

Integrated framework linking robust, data-efficient vision system methods with practical applications in digital technology development

Produced by

claude-haiku-4-5-20251001, under kortex-publication-v1