Appearance

Quick Start Guide ​

Welcome to Rosepetal AI Vision Platform ​

This guide will take you step by step from first login to making your first AI prediction in less than 30 minutes.

✅ Prerequisites ​

What You Need ​

  • Platform access: URL and credentials provided by your administrator
  • Web browser: Chrome, Firefox, Safari or Edge (recent versions)
  • Sample images: 20-50 images from your use case
  • Internet connection: Stable for uploading data

Supported Formats ​

  • Images: JPG, PNG, BMP, TIFF
  • Recommended size: 224x224 to 1024x1024 pixels
  • Maximum weight: 10MB per image

🚀 Step 1: First Access ​

Log In ​

  1. Open your browser and go to the Rosepetal URL
  2. Enter credentials:
    • Email: your email address
    • Password: provided by administrator
  3. Check "Remember" (optional) for longer sessions
  4. Click "Log in"

First Look at Dashboard ​

After login you'll see:

  • 🏠 Central dashboard with available modules
  • 👤 Your profile in the top right corner
  • 🔄 Language selector (Spanish/English)
  • 🌓 Dark/light mode available

💡 Tip: Take a few minutes to explore the interface and familiarize yourself with navigation.

📁 Step 2: Create Your First Dataset ​

New Dataset ​

  1. Click "Datasets" from the main dashboard

  2. Green "+" button to create new dataset

  3. Fill in information:

    • Name: "My_First_Dataset"
    • Description: Briefly describe the purpose
    • Type: Select according to your need:
      • 🏷️ Classification: To categorize (good/bad)
      • 🎯 Detection: To locate objects
      • 🎨 Segmentation: To delimit precise areas
      • 🚨 Anomalies: To detect defects

  4. Create dataset - it will open automatically

Upload Images ​

  1. "Upload" tab in the dataset view
  2. Drag and drop images or click to select
  3. Wait for complete upload - you'll see progress bar
  4. Verify that all images were uploaded correctly

📸 Recommendation: Start with 20-30 images for your first experiment.

🏷️ Step 3: Label Images ​

Access Labeling ​

  1. "Labeling" tab in your dataset
  2. The labeling editor will open in full screen

Label According to Type ​

For Classification ​

  1. View image in central canvas
  2. Select category from dropdown
  3. Next image automatically (→ arrow or button)
  4. Repeat for all images

For Object Detection ​

  1. Bounding box tool in sidebar
  2. Click and drag to create rectangle around object
  3. Select object class in popup
  4. Repeat for all objects in image
  5. Next image

For Segmentation ​

  1. Brush tool to paint areas
  2. Adjust brush size according to need
  3. Select class for painted area
  4. Eraser to correct errors
  5. Next image

Useful Shortcuts ​

  • ←→ Arrows: Navigate between images
  • Space: Zoom fit
  • +/-: Zoom in/out
  • Ctrl+Z: Undo

⚡ Speed Tip: Use keyboard shortcuts to label faster.

🤖 Step 4: Train Your First Model ​

Verify Dataset Ready ​

Before training, make sure:

  • ✅ All images labeled
  • ✅ Reasonable balance between classes
  • ✅ Minimum 20 images per class (50+ recommended)

Start Training ​

  1. "Models" tab in your dataset
  2. "Train Model" button or similar
  3. Configure basic parameters:
    • Model name: "My_First_Model_v1"
    • Description: Brief description of purpose
    • Architecture: Leave default initially
  4. Start training
Epochs: 50 (for simple classification)
Learning Rate: 0.001
Batch Size: 16
Data Augmentation: Enabled

Monitor Progress ​

  • Progress bar shows advancement
  • Real-time metrics: Loss, Accuracy
  • Estimated time remaining
  • Possibility to cancel if necessary

⏰ Estimated Time: 15-30 minutes for a small dataset.

📊 Step 5: Evaluate Results ​

View Metrics ​

Once training is completed:

  1. Final metrics are shown automatically
  2. Accuracy: Percentage of hits
  3. Training graphs: Metric evolution
  4. Confusion matrix: To understand errors (classification)

Interpret Results ​

Healthy Metrics ​

  • Accuracy > 80%: Good starting point
  • Decreasing loss: Model is learning
  • Validation similar to training: No overfitting

Warning Signs ​

  • Accuracy < 60%: Review data quality
  • Loss not decreasing: Possible configuration problem
  • Large train/validation difference: Possible overfitting

🎯 Step 6: Make Predictions ​

Test the Model ​

  1. In the trained model view
  2. "Test" or "Predict" button
  3. Upload new image (not used in training)
  4. View results:
    • Prediction: Class or detection
    • Confidence: Percentage of certainty
    • Visualization: Bounding boxes or masks

Interpret Predictions ​

For Classification ​

Result: "Defective"
Confidence: 92%

For Detection ​

Object 1: "Screw" (x:100, y:50, confidence: 88%)
Object 2: "Nut" (x:200, y:80, confidence: 91%)

For Segmentation ​

Defect area: 145 pixels (2.3% of image)
Average confidence: 87%

🎉 Congratulations! ​

You have completed your first complete project with Rosepetal:

  • ✅ Dataset created and populated
  • ✅ Images labeled
  • ✅ Model trained
  • ✅ Predictions made

🚀 Next Steps ​

Improve Your Model ​

  1. More data: Add more varied images
  2. Better labeling: Review and correct annotations
  3. Advanced parameters: Experiment with configurations
  4. External validation: Test with completely new data

Explore Advanced Features ​

  • 🔗 Flows: Automate processes with Node-RED
  • ⚙️ Controller: Monitor system and devices
  • 📺 Panels: Real-time control
  • 🔧 Settings: Advanced configurations

Integration and Production ​

  • API calls: Integrate predictions in other systems
  • Webhooks: Automate notifications
  • Batch processing: Process multiple images
  • Performance monitoring: Track performance in real use

💡 Final Tips ​

To Get Better Results ​

  • Quality before quantity: Better data = better model
  • Consistency: Uniform labeling criteria
  • Iteration: Continuous improvement with more data
  • External validation: Always test with new data

Avoid Common Errors ​

  • ❌ Too little data: Minimum 50 images per class
  • ❌ Inconsistent labeling: Define clear criteria
  • ❌ Not validating: Always test with new data
  • ❌ Overfitting: Watch validation metrics

📚 Additional Resources ​

Documentation ​

Help ​

Enjoy exploring the capabilities of Rosepetal AI Vision Platform! 🚀