← Back to curriculum

Module 7 — CV production & deployment

Monitoring, drift & retraining

Input distribution shift, concept drift, shadow deployments, human-in-the-loop labeling, and when to retrain vs fine-tune.

~65 min read + exercises

Monitoring, drift & retraining

Before we begin

Models degrade when reality diverges from training data. Production CV needs telemetry, alerts, and a retraining playbook.


Learning objectives

  • Distinguish input drift vs concept drift.
  • Monitor confidence, latency, and input statistics.
  • Use shadow deployments before cutover.
  • Decide fine-tune vs full retrain.

Input distribution shift

Examples: new camera sensor, seasonal lighting, different geographic region.

Monitor histograms of brightness, blur estimates, aspect ratios — alert when KL divergence or simple z-scores exceed thresholds.


Concept drift

Same pixels, different meaning — new product SKU, changed defect types. Labels from training no longer match reality.

Requires new labeled data and model update — monitoring accuracy proxies (human review sample) is essential.


Shadow deployment

Run candidate model on live traffic in parallel; log disagreements with production model — no user impact until validated.


Human-in-the-loop

Route low-confidence predictions to reviewers; corrected labels feed next training round (active learning).


Retraining policy

SignalAction
Small drift, backbone still validFine-tune on new batch
Large domain changeNew data collection + full pipeline review
Latency regressionProfile ONNX/TRT; consider smaller model

What's next

Module 7 quiz — then deploy vision API project.