Results

Matplotlib-like API for creating plots and visualizations displayed in real-time

🚧

Under Construction

This documentation is currently being developed and will be available soon.

Overview

The Results panel provides a Matplotlib-like API in the backend for creating plots and visualizations. Results are sent from Python via WebSocket and displayed in real-time in the frontend Results panel with version control, tagging, and interactive viewing features.

Backend API

ResultVisualizer Methods

from rfdt import server

# Plotting methods
server.results.plot(x, y, title="", xlabel="", ylabel="")      # Line plot
server.results.bar(x, heights, title="", xlabel="", ylabel="")  # Bar chart
server.results.scatter(x, y, title="", xlabel="", ylabel="")    # Scatter plot
server.results.imshow(data, title="", xlabel="", ylabel="")     # Heatmap/Image
server.results.table(data, columns=[], title="")                # Data table

# Utility methods
server.results.clear()                                          # Clear pending plots
server.results.commit(message="", tag="", tag_color="")        # Send to frontend
server.results.delete_version(version)                          # Delete a version
server.results.get_all_versions()                              # Get history

Plot Types

Line Plot

import numpy as np

x = np.linspace(0, 10, 100)
y = np.sin(x)
server.results.plot(x, y, title="Sine Wave", xlabel="Time", ylabel="Amplitude")

Bar Chart

categories = ['A', 'B', 'C', 'D']
values = [45, 38, 52, 41]
server.results.bar(categories, values, title="Comparison")

Scatter Plot

x = np.random.randn(100)
y = 2*x + np.random.randn(100)
server.results.scatter(x, y, title="Correlation")

Heatmap/Image

# Heatmap
data = np.random.randn(20, 30)
server.results.imshow(data, title="2D Field")

# RGB Image (H x W x 3)
image = np.random.rand(100, 100, 3) * 255
server.results.imshow(image.astype(np.uint8), title="RGB Image")

Data Table

table_data = {
    'Model': ['ResNet', 'VGG', 'MobileNet'],
    'Accuracy': [0.945, 0.923, 0.887],
    'Speed': [32, 28, 45]
}
server.results.table(table_data, title="Model Comparison")

Version Management

Commit System

# Simple commit
server.results.commit()  # Creates "v1"

# With message
server.results.commit(message="Baseline Results")

# With tag and color
server.results.commit(
    message="Best Model",
    tag="best",
    tag_color="#00ff00",
    clear_after=True  # Clear plots after commit (default)
)

Version History Features

•Persistent: Survives across sessions and frontend refresh
•Ordering: Reverse chronological order (newest first)
•Deletable: Delete with confirmation modal
•Backend Storage: Unlimited version history

Tag System

Tags provide visual categorization for results with automatic or custom colors:

•Auto Color: Generated from tag name
•Custom Colors: Hex format (#RRGGBB)
•Common Tags: train, val, test, baseline, exp, best

trainvaltestbaselineexpbest

Data Handling

Automatic Type Conversion

The Results API automatically converts various data types:

• PyTorch tensors → Python lists
• NumPy arrays → Python lists
• Pandas DataFrames → Dict format
• Native Python lists/tuples

Examples

# PyTorch tensors
import torch
x = torch.linspace(0, 10, 100)
y = torch.sin(x)
server.results.plot(x, y)  # Auto-converted

# NumPy arrays
import numpy as np
x = np.linspace(0, 10, 100)
y = np.sin(x)
server.results.plot(x, y)  # Auto-converted

Usage Workflows

Machine Learning Training

for epoch in range(epochs):
    train_model()

    # Clear previous plots
    server.results.clear()

    # Plot metrics
    server.results.plot(loss_history, title="Training Loss")
    server.results.bar(['Acc', 'Prec', 'Recall'], [acc, prec, recall])

    # Commit with tag
    server.results.commit(
        message=f"Epoch {epoch+1}",
        tag="train" if training else "val"
    )

Data Analysis

# Visualize distributions
server.results.scatter(data.x, data.y, title="Distribution")
server.results.imshow(correlation_matrix, title="Correlations")

# Show statistics
stats_table = compute_statistics(data)
server.results.table(stats_table, title="Summary")

server.results.commit(message="Analysis Complete", tag="analysis")

Frontend Features

Panel Features

• Version list sidebar
• Dynamic plot rendering
• Fullscreen mode
• Size adjustment slider (50%-200%)
• Delete with confirmation

UI Components

• Version List: Compact sidebar with tags
• Plot Area: Flexible grid layout
• Controls: Size slider, fullscreen button
• Delete Modal: Confirmation dialog

Performance Guidelines

•Max plots per version: ~50 plots
•Max data points per plot: ~10,000 points
•Version history: Unlimited (backend stored)

Troubleshooting

Issue	Solution
Plots not appearing	Check WebSocket connection and ensure commit() called
Black images	Ensure RGB data is uint8 with values 0-255
Tags not colored	Use valid hex format (#RRGGBB)
Delete not working	Check WebSocket connection status