Pat MulleeAI Product & Platform Leadership
Skip to primary interaction
Product Build

Computer Vision: Build or Buy?

Try six browser-based vision capabilities and examine deployment, cost, and privacy tradeoffs.

  • Working examples of six CV primitives running client-side
  • Build vs buy vs fine-tune cost framework per capability
  • Privacy by design, no image data leaves the browser
Try object detectionBuild vs. buy framework

Primary interaction

TensorFlow.js, COCO-SSD

Object Detection

Identify and locate objects in an image across ~80 everyday categories, people, vehicles, animals, household items.

Sample workspace scene for object detection, laptop, monitor, plant, mouse
Detecting objects…
Build vs. buy decision

Browser for ≤80 generic classes and privacy-first. Cloud for broad-vocabulary or higher accuracy without ML staff. Fine-tune for domain shift.

  • Inventory counting from a phone camera
  • Loss-prevention prefiltering in retail
  • Smart cropping and auto-tagging for content tools

Run the capability first. Decide how to ship it second.

  1. 01

    Start with the off-the-shelf model

    Prove the use case before funding custom training.

  2. 02

    Inspect the failure mode

    Where the model fails matters more than one aggregate accuracy number.

  3. 03

    Apply privacy and latency constraints

    On-device requirements may remove cloud choices immediately.

  4. 04

    Then choose browser, cloud, or custom

    Fine-tuning is the last move, not the first.