Quick Start Guide

Get started with computer vision in minutes. No technical knowledge required.

Step 1: Choose Your Platform

For this quick start: We'll walk through the basic workflow. The same steps work across all platforms.

Step 2: Create Your Project

Give your project a name—something descriptive like "Safety Equipment Detector" or "Quality Control". The system will organize everything for you automatically.

Simple, fast, and intuitive.

Step 3: Collect Images

Capture images of what you want to detect. Whether it's safety equipment, products, defects, or anything else—if you can photograph it, we can work with it.

Start with just a few images to test things out. Variety helps—different angles, lighting, and conditions give better results.

Step 4: Label Your Images

Now tell the system what to look for. It's intuitive—just draw boxes around objects and tag them. The interface is designed to be simple and fast, whether you're on a touchscreen or using a mouse.

Create labels for what you want to detect (like "hard hat" or "defect"), then draw boxes and assign them. Move through your images quickly and efficiently.

Step 5: Start Training

Once you've labeled your images, simply hit the "Train" button. The system handles everything automatically—choosing the right settings, optimizing for accuracy, and building your computer vision model.

You'll get a notification when it's ready. Most models complete in under an hour.

Step 6: Test Your Model

Once training completes:

1. Open the notification
2. Review accuracy metrics:
   • Overall accuracy: % of correct predictions
   • Per-class performance: How well it detects each class
   • Sample predictions: See it working on test images
3. Tap "Try It Live" to test with your camera
4. Point camera at objects - see real-time predictions!

What's Next?

You now have working computer vision!

Here's what you can do next:

Making Your Model Better

If accuracy isn't good enough (below 85%):

1. Add more images: Aim for 50-100 per class
2. Label edge cases: Include difficult examples (poor lighting, partial views, etc.)
3. Balance your classes: Make sure each class has similar numbers of examples
4. Review mistakes: Look at the confusion matrix to see what the model confuses
5. Retrain: With more data, accuracy improves dramatically

Try These Sample Projects

Want to explore before committing? Try our pre-built samples:

Safety Equipment Detection

• Detects: Hard hats, safety vests, gloves, goggles
• 50 pre-labeled images
• Train and test in minutes

Product Recognition

• Identifies different product types
• 60 pre-labeled retail product images
• See how it works for inventory

Quality Control

• Detects defects in manufactured parts
• 40 good + 40 defective examples
• Experience binary classification

Load any sample project from the app's main menu: "Load Sample Project"

Common Questions

How many images do I need?

Minimum: 10 per class to train
Recommended: 50-100 per class for good accuracy
More is better: You can always add more and retrain

Can I use existing photos?

Yes! Upload from your computer, Google Drive, or import from your camera roll.

Do I need internet?

For labeling: No, the app works offline
For training: Yes, training happens in the cloud
For using the model: No, models can run offline on your device

How much does it cost?

We offer a free tier to get started and explore the platform. Paid plans provide additional capacity and team features for scaling up. Contact us for details that fit your needs.

What if I need help?

We're here for you:

• Email: team@mlpractitioners.com
• Documentation: Detailed guides for every feature

Learn More