evlib 0.4.3

Python bindings for event camera utilities

evlib: Event Camera Data Processing Library

An event camera processing library with Rust backend and Python bindings, designed for scalable data processing with real-world event camera datasets.

Core Features

• Universal Format Support: Load data from H5, AEDAT, EVT2/3, AER, and text formats
• Automatic Format Detection: No need to specify format types manually
• Polars DataFrame Integration: High-performance DataFrame operations with up to 360M events/s filtering
• Event Filtering: Comprehensive filtering with temporal, spatial, and polarity options
• Event Representations: Stacked histograms, voxel grids, and mixed density stacks
• Neural Network Models: E2VID model loading and inference
• Real-time Data Processing: Handle large datasets (550MB+ files) efficiently
• Polarity Encoding: Automatic conversion between 0/1 and -1/1 polarities
• Rust Performance: Memory-safe, high-performance backend with Python bindings

In Development: Advanced neural network processing (hopefully with Rust backend, maybe Candle) Real-time visualization (Only simulated working at the moment — see wasm-evlib)

Note: The Rust backend currently focuses on data loading and processing, with Python modules providing advanced features like filtering and representations.

---

• Quick Start
  • Basic Usage
  • Advanced Filtering
  • Event Representations
• Installation
  • Basic Installation
  • Development Installation
  • System Dependencies
  • Performance-Optimized Installation
• Polars DataFrame Integration
  • Key Benefits
  • API Overview
    • Loading Data
    • Advanced Features
    • Utility Functions
  • Performance Benchmarks
  • Benchmarking and Monitoring
  • Performance Examples
    • Optimal Loading for Different File Sizes
    • Memory Monitoring
  • Troubleshooting Large Files
    • Memory Constraints
    • Performance Tuning
    • Common Issues and Solutions
• Available Python Modules
  • Core Modules
  • Module Overview
• Examples
• Development
  • Testing
    • Core Testing
    • Documentation Testing
    • Code Quality
  • Building
    • Requirements
• Community & Support
• License

Quick Start

Basic Usage

import evlib

# Load events from any supported format (automatic detection)
df = evlib.load_events("data/slider_depth/events.txt").collect(engine='streaming')

# Or load as LazyFrame for memory-efficient processing
lf = evlib.load_events("data/slider_depth/events.txt")

# Basic event information
print(f"Loaded {len(df)} events")
print(f"Resolution: {df['x'].max()} x {df['y'].max()}")
print(f"Duration: {df['timestamp'].max() - df['timestamp'].min()}")

# Convert to NumPy arrays for compatibility
x_coords = df['x'].to_numpy()
y_coords = df['y'].to_numpy()
timestamps = df['timestamp'].to_numpy()
polarities = df['polarity'].to_numpy()

Advanced Filtering

import evlib
import evlib.filtering as evf

# High-level preprocessing pipeline
processed = evf.preprocess_events(
    "data/slider_depth/events.txt",
    t_start=0.1, t_end=0.5,
    roi=(100, 500, 100, 400),
    polarity=1,
    remove_hot_pixels=True,
    remove_noise=True,
    hot_pixel_threshold=99.9,
    refractory_period_us=1000
)

# Individual filters (work with LazyFrames)
events = evlib.load_events("data/slider_depth/events.txt")
time_filtered = evf.filter_by_time(events, t_start=0.1, t_end=0.5)
spatial_filtered = evf.filter_by_roi(time_filtered, x_min=100, x_max=500, y_min=100, y_max=400)
clean_events = evf.filter_hot_pixels(spatial_filtered, threshold_percentile=99.9)
denoised = evf.filter_noise(clean_events, method="refractory", refractory_period_us=1000)

Event Representations

import evlib
import evlib.representations as evr

# TODO: Representation functions have Polars compatibility issues
# Create stacked histogram (replaces RVT preprocessing)
# events = evlib.load_events("data/slider_depth/events.txt")
# events_df = events.collect()
# hist = evr.create_stacked_histogram_py(
#     events_df,
#     _height=480, _width=640,
#     nbins=10, window_duration_ms=50.0,
#     _count_cutoff=10
# )

# Create mixed density stack representation
# density = evr.create_mixed_density_stack_py(
#     events_df,
#     _height=480, _width=640,
#     nbins=10, window_duration_ms=50.0
# )

# TODO: High-level preprocessing for neural networks
# data = evr.preprocess_for_detection(
#     "data/slider_depth/events.txt",
#     representation="stacked_histogram",
#     height=480, width=640,
#     nbins=10, window_duration_ms=50
# )

# TODO: Performance benchmarking against RVT
# results = evr.benchmark_vs_rvt("data/slider_depth/events.txt", height=480, width=640)

Installation

Basic Installation

pip install evlib

# For Polars DataFrame support (recommended)
pip install evlib[polars]

Development Installation

# Clone the repository
git clone https://github.com/tallamjr/evlib.git
cd evlib

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install in development mode with all features
pip install -e ".[dev,polars]"

# Build the Rust extensions
maturin develop

System Dependencies

# Ubuntu/Debian
sudo apt install libhdf5-dev pkg-config

# macOS
brew install hdf5 pkg-config

Performance-Optimized Installation

For optimal performance, ensure you have the recommended system configuration:

System Requirements:

• RAM: 8GB+ recommended for files >100M events
• Python: 3.10+ (3.12 recommended for best performance)
• Polars: Latest version for advanced DataFrame operations

Installation for Performance:

# Install with Polars support (recommended)
pip install "evlib[polars]"

# For development with all performance features
pip install "evlib[dev,polars]"

# Verify installation with benchmark
python -c "import evlib; print('evlib installed successfully')"
python benchmark_memory.py  # Test memory efficiency

Optional Performance Dependencies:

# For advanced memory monitoring
pip install psutil

# For parallel processing (already included in dev)
pip install multiprocessing-logging

Polars DataFrame Integration

evlib provides comprehensive Polars DataFrame support for high-performance event data processing:

Key Benefits

• Performance: 1.9M+ events/s loading speed, 360M+ events/s filtering speed
• Memory Efficiency: ~23 bytes/event (5x better than typical 110 bytes/event)
• Expressive Queries: SQL-like operations for complex data analysis
• Lazy Evaluation: Query optimization for better performance
• Ecosystem Integration: Seamless integration with data science tools

API Overview

Loading Data

import evlib

# Load as LazyFrame (recommended)
events = evlib.load_events("data/slider_depth/events.txt")
df = events.collect()  # Collect to DataFrame when needed

# Automatic format detection and optimization
events = evlib.load_events("data/slider_depth/events.txt")  # EVT2 format automatically detected
print(f"Format: {evlib.formats.detect_format('data/slider_depth/events.txt')}")
print(f"Description: {evlib.formats.get_format_description('EVT2')}")

Advanced Features

import evlib
import polars as pl

# Chain operations with LazyFrames for optimal performance
events = evlib.load_events("data/slider_depth/events.txt")
result = events.filter(pl.col("polarity") == 1).with_columns([
    pl.col("timestamp").dt.total_microseconds().alias("time_us"),
    (pl.col("x") + pl.col("y")).alias("diagonal_pos")
]).collect()

# Memory-efficient temporal analysis
time_stats = events.with_columns([
    pl.col("timestamp").dt.total_microseconds().alias("time_us")
]).group_by([
    (pl.col("time_us") // 1_000_000).alias("time_second")  # Group by second
]).agg([
    pl.len().alias("event_count"),
    pl.col("polarity").mean().alias("avg_polarity")
]).collect()

# Combine with filtering module for complex operations
import evlib.filtering as evf
filtered = evf.filter_by_time(events, t_start=0.1, t_end=0.5)
analysis = filtered.with_columns([
    pl.col("timestamp").dt.total_microseconds().alias("time_us")
]).collect()

Utility Functions

import evlib
import polars as pl
import evlib.filtering as evf

# Built-in format detection
format_info = evlib.formats.detect_format("data/slider_depth/events.txt")
print(f"Detected format: {format_info}")

# Spatial filtering using dedicated filtering functions (preferred)
events = evlib.load_events("data/slider_depth/events.txt")
spatial_filtered = evf.filter_by_roi(events, x_min=100, x_max=200, y_min=50, y_max=150)

# Or using direct Polars operations
manual_filtered = events.filter(
    (pl.col("x") >= 100) & (pl.col("x") <= 200) &
    (pl.col("y") >= 50) & (pl.col("y") <= 150)
)

# Temporal analysis with Polars operations
rates = events.with_columns([
    pl.col("timestamp").dt.total_microseconds().alias("time_us")
]).group_by([
    (pl.col("time_us") // 10_000).alias("time_10ms")  # Group by 10ms
]).agg([
    pl.len().alias("event_rate"),
    pl.col("polarity").mean().alias("avg_polarity")
]).collect()

# TODO: Save functions have type compatibility issues
# Save processed data
# processed = evf.preprocess_events("data/slider_depth/events.txt", t_start=0.1, t_end=0.5)
# processed_df = processed.collect()
# x, y, t_us, p = processed_df.select(["x", "y", "timestamp", "polarity"]).to_numpy().T
# Convert microseconds to seconds for save function
# t = t_us.astype('float64') / 1_000_000
# evlib.formats.save_events_to_hdf5(x, y, t, p, "output.h5")

Performance Benchmarks

Benchmark Results:

• Loading Speed: 1.9M+ events/second average across formats
• Filter Speed: 360M+ events/second for complex filtering operations
• Memory Efficiency: ~23 bytes/event
• Format Performance: RAW binary (2.6M events/s) > HDF5 (2.5M events/s) > Text (0.6M events/s)

Benchmarking and Monitoring

Run performance benchmarks to verify optimizations:

# Verify README performance claims and generate plots
python benches/benchmark_performance_readme.py

# Memory efficiency benchmark
python benches/benchmark_memory.py

# Test with your own data
python -c "
import evlib
import time
start = time.time()
events = evlib.load_events('data/slider_depth/events.txt')
df = events.collect()
print(f'Loaded {len(df):,} events in {time.time()-start:.2f}s')
print(f'Format: {evlib.detect_format(\"data/slider_depth/events.txt\")}')
print(f'Memory per event: {df.estimated_size() / len(df):.1f} bytes')
"

Performance Examples

Optimal Loading for Different File Sizes

import evlib
import evlib.filtering as evf
import polars as pl

# Small files (<5M events) - Direct loading
events_small = evlib.load_events("data/slider_depth/events.txt")
df_small = events_small.collect()

# Large files (>5M events) - Automatic streaming
events_large = evlib.load_events("data/slider_depth/events.txt")
# Same API, automatically uses streaming for memory efficiency

# Memory-efficient filtering on large datasets using filtering module
filtered = evf.filter_by_time(events_large, t_start=1.0, t_end=2.0)
positive_events = evf.filter_by_polarity(filtered, polarity=1)

# Or using direct Polars operations
manual_filtered = events_large.filter(
    (pl.col("timestamp").dt.total_microseconds() / 1_000_000 > 1.0) &
    (pl.col("polarity") == 1)
).collect()

Memory Monitoring

import evlib
import psutil
import os

def monitor_memory():
    process = psutil.Process(os.getpid())
    return process.memory_info().rss / 1024 / 1024  # MB

# Monitor memory usage during loading
initial_mem = monitor_memory()
events = evlib.load_events("data/slider_depth/events.txt")
df = events.collect()
peak_mem = monitor_memory()

print(f"Memory used: {peak_mem - initial_mem:.1f} MB")
print(f"Memory per event: {(peak_mem - initial_mem) * 1024 * 1024 / len(df):.1f} bytes")
print(f"Polars DataFrame size: {df.estimated_size() / 1024 / 1024:.1f} MB")

Troubleshooting Large Files

Memory Constraints

• Automatic Streaming: Files >5M events use streaming by default
• LazyFrame Operations: Memory-efficient processing without full materialization
• Memory Monitoring: Use benchmark_memory.py to track usage
• System Requirements: Recommend 8GB+ RAM for files >100M events

Performance Tuning

• Optimal Chunk Size: System automatically calculates based on available memory
• LazyFrame Operations: Use .lazy() for complex filtering chains
• Memory-Efficient Formats: RAW binary formats provide best performance, followed by HDF5
• Progress Reporting: Large files show progress during loading

Common Issues and Solutions

Issue: Out of memory errors

import evlib
import evlib.filtering as evf

# Solution: Use filtering before collecting (streaming activates automatically)
events = evlib.load_events("data/slider_depth/events.txt")
# Streaming activates automatically for files >5M events

# Apply filtering before collecting to reduce memory usage
filtered = evf.filter_by_time(events, t_start=0.1, t_end=0.5)
df = filtered.collect()  # Only collect when needed

# Or stream to disk using Polars
filtered.sink_parquet("filtered_events.parquet")

Issue: Slow loading performance

import evlib
import evlib.filtering as evf
import polars as pl

# Solution: Use LazyFrame for complex operations and filtering module
events = evlib.load_events("data/slider_depth/events.txt")

# Use filtering module for optimized operations
result = evf.filter_by_roi(events, x_min=0, x_max=640, y_min=0, y_max=480)
df = result.collect()

# Or chain Polars operations
result = events.filter(pl.col("polarity") == 1).select(["x", "y", "timestamp"]).collect()

Issue: Memory usage higher than expected

import evlib

# Solution: Monitor and verify optimization
events = evlib.load_events("data/slider_depth/events.txt")
df = events.collect()
print(f"Memory efficiency: {df.estimated_size() / len(df)} bytes/event")
print(f"DataFrame schema: {df.schema}")
print(f"Number of events: {len(df):,}")

# Check format detection
format_info = evlib.formats.detect_format("data/slider_depth/events.txt")
print(f"Format: {format_info}")

Available Python Modules

evlib provides several Python modules for different aspects of event processing:

Core Modules

• evlib.formats: Direct Rust access for format loading and detection
• evlib.filtering: High-performance event filtering with Polars
• evlib.representations: Event representations (stacked histograms, voxel grids)
• evlib.models: Neural network model loading and inference (Under construction)

Module Overview

import evlib
import evlib.filtering as evf
import evlib.representations as evr

# Core event loading (returns Polars LazyFrame)
events = evlib.load_events("data/slider_depth/events.txt")

# Format detection and description
format_info = evlib.formats.detect_format("data/slider_depth/events.txt")
description = evlib.formats.get_format_description("HDF5")

# Advanced filtering
filtered = evf.preprocess_events("data/slider_depth/events.txt", t_start=0.1, t_end=0.5)
time_filtered = evf.filter_by_time(events, t_start=0.1, t_end=0.5)

# TODO: Event representations have Polars compatibility issues
# Event representations (need to load data first)
# events_df = events.collect()
# hist = evr.create_stacked_histogram_py(events_df, _height=480, _width=640, nbins=10)
# voxel = evr.create_voxel_grid_py(events_df, _height=480, _width=640, nbins=5)

# Neural network models (limited functionality)
from evlib.models import ModelConfig  # If available

# TODO: Data saving functions have type compatibility issues
# Data saving (need to get arrays first)
# df = events.collect()
# x, y, t_us, p = df.select(["x", "y", "timestamp", "polarity"]).to_numpy().T
# Convert microseconds to seconds for save functions
# t = t_us.astype('float64') / 1_000_000
# evlib.formats.save_events_to_hdf5(x, y, t, p, "output.h5")
# evlib.formats.save_events_to_text(x, y, t, p, "output.txt")

Examples

Run examples:

# Test all notebooks
pytest --nbmake examples/

# Run specific examples
python examples/simple_example.py
python examples/filtering_demo.py
python examples/stacked_histogram_demo.py

Development

Testing

Core Testing

# Run all tests (Python and Rust)
pytest
cargo test

# Test specific modules
pytest tests/test_filtering.py
pytest tests/test_representations.py
pytest tests/test_evlib_exact_match.py

# Test notebooks (including examples)
pytest --nbmake examples/

# Test with coverage
pytest --cov=evlib

Documentation Testing

All code examples in the documentation are automatically tested to ensure they work correctly:

# Test all documentation examples
pytest --markdown-docs docs/

# Test specific documentation file
pytest --markdown-docs docs/getting-started/quickstart.md

# Use the convenient test script
python scripts/test_docs.py --list    # List testable files
python scripts/test_docs.py --report  # Generate report

# Test specific documentation section
pytest --markdown-docs docs/user-guide/
pytest --markdown-docs docs/getting-started/

Code Quality

# Format code
black python/ tests/ examples/
cargo fmt

# Run linting
ruff check python/ tests/
cargo clippy

# Check types
mypy python/evlib/

Building

Requirements

• Rust: Stable toolchain (see rust-toolchain.toml)
• Python: ≥3.10 (3.12 recommended)
• Maturin: For building Python extensions

# Development build
maturin develop --features python # python required to register python modules

# Build with features
maturin develop --features polars
maturin develop --features pytorch

# Release build
maturin build --release

Community & Support

{ width=100% }

• GitHub: tallamjr/evlib
• Issues: Report bugs and request features
• Discussions: Community Q&A and ideas

License

MIT License - see LICENSE.md for details.