ChiefOSby ConversintStart free
Menu
← Back to the library
Data science and analytics

Computer vision: bigru slice feature aggregator

Two-stage CT classifier where a 2D CNN dumps per-slice features once, then a bidirectional GRU runs over the slice sequence to produce both per-slice predictions (TimeDistributed head) and an exam-level prediction (avg+max pooled head) — turns expensive 3D CNN training into cheap sequence modeling

Works with the context you provideVersion 1.0.0

Cv bigru slice feature aggregator

Use this skill for the outcome in its description. Start with the supplied request and evidence, identify the specific task, and distinguish planning from execution. Do not trigger solely because the subject happens to be fashion, retail, or a technology named in an example. Read `references/execution-contract.md` before applying a retained source method.

Select the narrowest matching method below, then read its full entrypoint and the resources it requires. If multiple methods cover the same outcome, choose the one matching the user's environment and inputs; do not run them all. Retain each method's input requirements, decision steps, validation, output structure, and failure handling. If prerequisites are absent, use its honest proposal-only fallback or state that its documented path is blocked.

  • **wenmin-wu/cv-bigru-slice-feature-aggregator**: Two-stage CT classifier where a 2D CNN dumps per-slice features once, then a bidirectional GRU runs over the slice sequence to produce both per-slice predictions (TimeDistributed head) and an exam-level prediction (avg+max pooled head) — turns expensive 3D CNN training into cheap sequence modeling Read `references/wenmin-wu/skills/cv/bigru-slice-feature-aggregator/SKILL.md`.

Return the requested deliverable with evidence, assumptions, validation performed, unresolved dependencies, and actual execution status. Source-specific results must be verified independently; importing these methods does not establish a working service connection.