| name | vss-generate-video-calibration |
| description | Use this skill when running AutoMagicCalib on local MP4s, RTSP, or the bundled sample dataset, or when deploying vss-auto-calibration. Do not use for non-AMC calibration or runtime analytics. |
| license | Apache-2.0 |
| metadata | {"version":"3.3.0","github-url":"https://github.com/NVIDIA-AI-Blueprints/video-search-and-summarization","tags":"nvidia blueprint operational"} |
Purpose
Run AutoMagicCalib end-to-end on local files, RTSP streams, or the bundled sample dataset and (when needed) deploy the AMC microservice.
Instructions
Follow the routing tables and step-by-step workflows below. Each section that ends in workflow, quick start, or flow is intended to be executed top-to-bottom. Detailed reference material lives in references/; load only the reference needed for the selected input mode.
Examples
Worked end-to-end examples are kept under evals/ (each *.json manifest contains a runnable scenario) and inline in the per-workflow curl blocks below. Run a Tier-3 evaluation with nv-base validate <this-skill-dir> --agent-eval to replay them.
Limitations
- Requires the matching VSS profile / microservice to be deployed and reachable from the caller.
- NGC-hosted models and NIMs may be subject to rate-limits, GPU memory requirements, and license restrictions.
- Concurrency, GPU memory, and storage limits depend on the host hardware and the profile's compose file.
Troubleshooting
- Error: REST call returns connection refused. Cause: target microservice not running. Solution: probe
/docs or /health; redeploy via vss-deploy-profile or the matching vss-deploy-* skill.
- Error: HTTP 401/403 from NGC pulls. Cause: missing/expired
NGC_CLI_API_KEY. Solution: docker login nvcr.io and re-export the key before retrying.
- Error: container OOM or model fails to load. Cause: insufficient GPU memory for the selected profile. Solution: switch to a smaller variant or free GPUs via
docker compose down.
VSS Generate Video Calibration
Run AutoMagicCalib over one of three input sources and drive the calibration through the microservice REST API. The input-resolution work differs per source; everything from verify_project onward is identical and lives in this file. Pick the right input-mode reference and pair it with the Shared Calibration Tail below.
Shared helper references are loaded only when needed:
- Read
references/common-steps.md when a mode reference needs the shared create_project, video-upload, or handoff snippets.
- Read
references/calibration-tail.md when you need the reusable Python implementation of the stage-linear-media → verify → VGGT/post-process → AMC/post-process → compare-results tail.
Input Routing
Match the user's request to a mode, then load that mode's reference for input collection, mode-specific API calls, and the full Python script.
Disambiguation rule: if the user is asking to launch / deploy / set up AMC (no calibration verb) → deploy. If they provide RTSP URLs → rtsp. If they mention local files / a videos directory → videos. If they ask to verify install or test the bundled sample → sample-dataset. Combined intents (e.g. "launch AMC and calibrate my videos") → walk deploy first, then the calibration mode. When ambiguous, ask via AskUserQuestion.
Prerequisites (shared across calibration modes)
- Platform preflight from
references/deploy-auto-calibration-service.md Step 0 passes before any AMC deploy or calibration API work. The calibration host needs Ubuntu 24.04 on x86_64, NVIDIA Driver 590 or newer, NVIDIA GPU access, and NVENC hardware encoder support. If the preflight fails, stop immediately, tell the user which requirement was not met, and ask them to provide an existing calibration.json, run calibration on a supported x86_64 dGPU host, or transfer generated calibration artifacts. Do not continue AMC setup, VIOS probing, capture, upload, or calibration automatically. DGX Spark is aarch64, so use existing/generated artifacts for this flow.
- AMC microservice + UI running. If not, walk
references/deploy-auto-calibration-service.md first.
- Microservice reachable at
http://<HOST_IP>:${VSS_AUTO_CALIBRATION_HOST_PORT:-8010}/v1/ready → {"code":0,...}.
- Projects directory writable by the container user. If you didn't just deploy (so Step 5 of the deploy reference hasn't run), confirm the write test in
references/deploy-auto-calibration-service.md § Step 5 — otherwise the first create_project returns [Errno 13] Permission denied.
- Python 3 with
requests installed (each input-mode reference includes a self-healing venv fallback for direct runs).
Mode-specific prerequisites (VIOS for rtsp, sample zip for sample-dataset) live in the respective references. The platform preflight applies even when an AMC service is already running.
Shared Calibration Tail
The shared sequence is stage-linear-media → verify → VGGT (when ready) → post-process → AMC → post-process → results. After the mode-specific reference has uploaded videos / automatically ingested RTSP clips / uploaded the bundled sample, run this tail. Use references/calibration-tail.md for the shared Python snippet.
AMC UI sequence: Step 1 Project Setup, Step 2 Video Configuration, Step 3 Parameters, Step 4 Rectification, Step 5 Manual Alignment, Step 6 Execute, Step 7 Results.
Step A — Stage Linear Media
AMC v3.3.0 cannot calibrate raw media. After the mode-specific workflow has uploaded videos or completed RTSP ingest, explicitly choose one path before verification:
- Already-linear/pinhole media — call
POST /v1/linear_media/<project_id> and require rectification_state == "COMPLETED".
- Distorted media — open AMC UI Step 4: Rectification; select Auto, Manual, or Videos Are Rectified; review the estimate; then click Generate Rectified Videos. Auto supports
simple_divisional (default), simple_radial, and radial; Manual supports per-camera model, k1, and k2 for radial. READY_FOR_REVIEW is not complete: require rectification_state == "COMPLETED" before continuing. Re-rectification invalidates verification, calibration, and post-processing outputs.
For REST-only rectification, use the running AMC service contract exposed by <MS_URL>/docs (OpenAPI: <MS_URL>/openapi.json):
- Auto —
POST /v1/rectification/<project_id> starts frame-0 estimation. Poll GET /v1/rectification/<project_id> until READY_FOR_REVIEW; then GET /v1/rectification/<project_id>/cameras, review every auto_estimate, and commit all camera parameters with POST /v1/rectification/<project_id>/manual using {"cameras":{"cam_00":{"model":"...","k1":0.0,"k2":0.0},...}}. This explicit commit generates full rectified videos.
- Manual —
POST /v1/rectification/<project_id>/manual/start; optionally preview each adjustment through POST /v1/rectification/<project_id>/preview/<camera_id>; then commit a complete per-camera cameras map to POST /v1/rectification/<project_id>/manual.
- Poll
GET /v1/rectification/<project_id> until COMPLETED. Stop on ERROR; do not verify or calibrate from READY_FOR_REVIEW.
Rectification produces rectified.mp4 and rectified.jpg. External alignment files normally use coord_space=original; use rectified only for points created on AMC rectified media. Never call /v1/calibrate/<project_id> before the linear-media or rectification state is complete.
Step B — Verify Project
POST /v1/verify_project/<project_id>
Response: {"project_state": "READY"} — must be READY before calibrating. If not READY, re-check that videos + alignment + layout are present (either via API or via UI manual alignment).
Step C — Independent VGGT Calibration
After verification and before AMC, inspect vggt_state. Start VGGT by default from READY, resume and wait from RUNNING, and always post-process a COMPLETED multi-camera result, including one completed before the current invocation. MODEL_MISSING or ERROR is reported as an AMC-only fallback. Check amc_state, vggt_state, and postprocess_state independently.
POST /v1/vggt/calibrate/<project_id>
GET /v1/get_project_info/<project_id> # poll vggt_state
POST /v1/postprocess/<project_id> # multi-camera only, after VGGT
GET /v1/get_project_info/<project_id> # require postprocess_state == COMPLETED
GET /v1/vggt_results/<project_id>/evaluation_statistics # VGGT metrics when GT exists
Step D — Start AMC Calibration
Confirm the plan before calibrating. Whether the settings file and detector were auto-detected or asked, present a short summary and confirm via AskUserQuestion before the POST /calibrate. The resolved values are the defaults, so confirming is one click — but the user can switch the detector or skip an auto-detected settings file. Summarize:
- Detector —
resnet or transformer (the value to be sent).
- Calibration settings — the file being applied (path), or default parameters (with the option to tune them in the UI first — see below).
- Optional overrides — ground-truth zip and focal lengths, if any.
The sample-dataset install-check run uses a fixed resnet and can proceed without this confirmation.
POST /v1/calibrate/<project_id>
Content-Type: application/json
{"detector_type": "resnet"} # or "transformer"
detector_type is a separate /calibrate parameter — not consumed by /v1/config/<id>. If the user provided a calibration settings file, parse it for "detector" / "detector_type" and use that value. If the file doesn't specify one, the default (resnet) is the value shown in the confirmation above — the user can switch it there before calibrating. If there's no settings file at all, ask the user via AskUserQuestion:
resnet — default, fast.transformer — slower, better under heavy occlusion.
UI Step 3 (Parameters) does NOT cover detector choice; never assume the user picked one in the UI.
Also when there's no settings file, ask whether to tune the calibration parameters first (AskUserQuestion):
- Proceed with the default parameters — well-suited to typical warehouse scenes; recommended unless the user has specific tuning in mind.
- Adjust parameters in the UI first — open the project, go to Step 3: Parameters, change values, and click Save; then continue.
In Step 3, set layout_px_per_m directly or measure a known two-point distance. Re-run post-processing after a scale or alignment change.
Wait for the user's choice — and, if they choose to tune, for them to confirm they've Saved — before calling /calibrate.
Step E — Poll for AMC Completion
GET /v1/get_project_info/<project_id>
Poll every 10 s. Use project_info.amc_state for AMC completion; aggregate project_state is not a pipeline-success signal.
| State | Meaning |
|---|
RUNNING | AMC calibration in progress |
COMPLETED | Finished |
ERROR | Failed — pull log via GET /v1/amc/calibrate/<id>/log |
When calibration starts, surface the project ID, the UI URL (http://<HOST_IP>:${VSS_AUTO_CALIBRATION_UI_HOST_PORT:-5000}), and the log endpoint so the user can watch progress while the run proceeds. During RUNNING, emit a progress line at least once a minute with elapsed time so a long run doesn't look stalled. On ERROR, fetch and show the last lines of GET /v1/amc/calibrate/<id>/log before stopping. Live logs can also be streamed via GET /v1/calibrate/<project_id>/log/<type>/stream.
Typical time: 10–60 min (your-own videos), 10–30 min (bundled sample). A six-camera transformer run can exceed one hour; keep polling and inspect logs/UI instead of treating 60 minutes as failure.
Step F — AMC Post-process and Results
For multi-camera projects, run layout post-processing after AMC calibration. VGGT, when available, runs first and is post-processed before AMC.
POST /v1/postprocess/<project_id>
GET /v1/get_project_info/<project_id> # poll postprocess_state until COMPLETED
Do not report a multi-camera project as successful until postprocess_state == "COMPLETED"; raw AMC results may exist even when post-processing fails.
GET /v1/get_project_info/<project_id> # project state
GET /v1/result/<project_id>/evaluation_statistics # only if GT uploaded
GET /v1/result/<project_id>/overlay_image # visual overlay (PNG)
GET /v1/amc/calibrate/<project_id>/log # calibration log
Evaluation response includes Average L2 distance(m) and Average reprojection error 0(px). Evaluation metrics are produced only when a ground-truth GT.zip was uploaded — a missing evaluation_statistics result is normal otherwise and is not the end of result reporting. When VGGT also completed, compare both methods' metrics and Results-page overlays, then select the more accurate calibration for export.
After COMPLETED, always give the user a way to review the result for that exact project, regardless of whether metrics exist:
- UI —
http://<HOST_IP>:${VSS_AUTO_CALIBRATION_UI_HOST_PORT:-5000}; open the project, then the Results page to view the overlay.
- Overlay image on disk —
${VSS_APPS_DIR}/services/auto-calibration/projects/project_<id>/output/multi_view_results/BA_output/results_ba_scaled_world/overlay_img_*.png (single-camera projects use output/single_view_results/cam_00/verification_map_overlay.png).
- Project files —
${VSS_APPS_DIR}/services/auto-calibration/projects/project_<id>/.
Settings File + Detector Pattern
Optional across all three modes. Before using a JSON settings file, retrieve GET /v1/config/defaults and inspect <MS_URL>/openapi.json (or <MS_URL>/docs) from the running AMC version. Parse the file, reject known-invalid legacy skip rather than silently translating it to skip_frame, then submit the JSON unchanged in meaning. Do not treat /config/defaults as a complete allow-list: the running API is the authoritative schema validator.
POST /v1/config/<project_id>
Content-Type: application/json
<parsed JSON object; submit as application/json>
The file replaces what the user would otherwise tune in UI Step 3 (parameters, bundle-adjustment, and evaluation knobs). Rectification is UI Step 4 and follows Step A. After a successful POST, also parse the file for "detector" / "detector_type" — if it's "resnet" or "transformer", use that value for the /calibrate call in Step D (detector is a separate API parameter, not consumed by /config).
Non-2xx is surfaced — do not silently fall back. Skip this call entirely if the user chose the UI-fallback path.
UI Fallback Pattern
When alignment / layout files aren't on disk, direct the user to the appropriate AMC UI step:
- Settings missing → "Open UI project
<project_id>, go to Step 3: Parameters, tune via the settings dialog (or accept defaults), click Save." Also: before the /calibrate call, ask the user via AskUserQuestion whether to use the resnet or transformer detector — Step 3 doesn't cover detector choice.
- Layout missing → "Open UI project
<project_id>, go to Step 2: Video Configuration, upload layout.png only (do NOT re-upload videos — they're already attached via API/RTSP), click Save."
- Alignment missing → "Open UI project
<project_id>, go to Step 5: Manual Alignment, either upload alignment_data.json or mark correspondence points on the layout, click Save."
Wait for user confirmation. For alignment/layout, verify on disk before continuing:
