← ClaudeAtlas

detection-skilllisted

Find faces, text (OCR, Japanese and English, including vertical Japanese) and objects in a video's frames or in an image, as measured boxes with time, pixel box, score and model hash, link them over time into redact-ready rectangles, measure shot boundaries (cuts and dissolves) over every frame, turn face tracks into a reframing path for 9:16, 1:1 or 4:5, say who is speaking and which camera shows them best as a multicam switch list, and flag personal and patient data (ids, dates of birth, names, phones, addresses, faces) on slides for a person to confirm before it is blurred, and rank frames for a thumbnail by eyes open, sharpness and face angle. Use when asked to find (and then blur) faces, name badges, screens or other things in event footage, read text on slides, screenshots or signs, check where people or objects are in a shot, list a video's shots and cuts, plan a vertical / square reframe that follows the speaker, cut a multicam panel or interview to whoever is talking, check slides for patient data be
zobaerio/zobvia · ★ 1 · AI & Automation · score 74
Install: claude install-skill zobaerio/zobvia
> **Bundled with Zobvia** (extras/detection-skill/) — see the sub-skill parts of Zobvia's SKILL.md. Source: [kajisho5/detection-skill](https://github.com/kajisho5/detection-skill) v0.2.3 (main @ 43cc556), MIT License (LICENSE in this folder). Bundled: SKILL.md, `src/`, `schemas/`, `pyproject.toml`. Not bundled: docs, tests, brand. It reports boxes in time (faces/text/objects) — never identities. # detection-skill Run `detection <command> ... --json` (installed by `pip install -e .`; engines by `pip install -e '.[faces]'`, `'.[objects]'`, and the `tesseract` binary for text). Every command prints one JSON document. `detection doctor --json` says which engines are ready and, if not, what to install and where to get the model files (it never downloads anything). | Task | Command | |---|---| | Faces once a second | `detection faces talk.mp4 --every 1 -o faces.json --json` | | Faces at chosen times / once per scene | `--at 12.5,40` / `--scenes` | | Text on a slide or screenshot | `detection text slide.png -o text.json --json` (default `--lang jpn+eng`) | | English only / vertical Japanese | `--lang eng` / `--vertical` | | Sparse text in a video frame (signs, badges) | `detection text hall.mp4 --every 2 --psm 11 -o text.json` | | People and laptops | `detection objects hall.mp4 --every 2 --classes person,laptop -o objects.json --json` | | Text with PP-OCRv6 (optional) | `detection text slide.png --engine ppocrv6 -o text.json --json` (only when `doctor --json` shows `optional_re