Build your complete Event-Based Machine Learning pipelines with the Metavision® Intelligence Suite
Leveraging the extensive and real-sequence datasets that Prophesee has created over the past four years, use a wide variety of ML tools to guide the development of neural network models, run inference on Event-Based data for object detection and tracking, as well as optical flow estimation and classification – all optimized for Event-Based Vision.
Easily Train your own Event-Based Neural Models with or without Event-based data, thanks to our tailor-made training pipeline and Video to Event simulator.
DOWNLOAD AND BUILD NOW
Our software toolkit is available on Windows and Ubuntu PC. Design and build your own Metavision Intelligence-based application, leveraging Prophesee Event-Based Metavision sensing technologies.

MACHINE LEARNING – WORLD’S LARGEST HD EVENT-BASED DATASET
Our model, spotlighted at NeurIPS, outperforms by a large margin feed-forward event-based architectures.
Comparable performance to RetinaNet-50 on 1 Mpx grayscale images, but with a 27% smaller model and superior generalization ability under challenging lightning condition.
25M Bounding Boxes
15H Recording
1MP Resolution
START WITH YOUR EXISTING IMAGE-BASED DATA
Bridge frame-based and event-based worlds with our Video to Event pipeline. Generate synthetic data to augment your dataset, and partially reuse existing references.
Off-the-shelf, ready-to-use Event simulator
EVENT TO VIDEO
Build grayscale images based on events with this neural network. Spare resources by facilitating automatic Event-based data annotation for tasks already solved in Frame-based.
Off-the-shelf, ready-to-use Event to video converter
Leverage existing frame-based neural networks to obtain labels
RUN YOUR DETECTOR
Unlock the potential of Event-Based machine learning, with a set of dedicated tools providing everything you need to start execution of Deep Neural Network (DNN) with events. Leverage our pretrained automotive model written in pytorch, and experiment live detection & tracking using our c++ pipeline. Use our comprehensive python library to design your own networks.
Pretrained network trained on a 15h and 23M labels automotive dataset
Live detection and tracking @100Hz
TRAIN YOUR MODEL
Train your own Object Detection application with our ready-to use training framework. Experiment with multiple pre-built Event-Based Tensor representations, and training network topology suited for event-based Data.
4 pre-built tensor representation
Automated HDF5 dataset generation
Comprehensive training toolbox, including Tailor-made preprocessing, DataLoader, NN architecture, visualization tool and more
PREDICT OPTICAL FLOW
Predict optical flow from Event-Based data leveraging our pretrained Flow Model, customized data loader and collections of loss function and visualization tools to set up your flow inference pipeline.
Self-supervised Flownet architectures
Lightweight model
TRAIN YOUR OPTICAL FLOW MODEL
No ground truth ? Leverage our self-supervised architecture. Train your Optical Flow application with our custom-built FlowNet training framework. Experiment with 4 pre-built Flow Networks tailor-made for your Event-based data.
Comprehensive flow training toolbox, including different network topologies, various loss functions and visualization mode
RUN YOUR CLASSIFIER
Run a live Rock Paper Scissors game with our pre-trained model or run a live classifier with your own event-based recordings.
Off-the-shelf, ready-to-use classifier
TRAIN YOUR CLASSIFICATION MODEL
Train your own object classifier with this tailor-made Event-based training pipeline.
Comprehensive training toolbox, including Tailor-made preprocessing, DataLoader, NN architecture, visualization tool and more
EVALUATE YOUR EVENT-BASED NETWORK
Evaluate your detection performance with our Object Detection KPI toolkit in line with the latest COCO API.
mAP, mAR and their variants included
DOWNLOAD AND BUILD NOW
Our software toolkit is available on Windows and Ubuntu PC. Design and build your own Metavision Intelligence-based application, leveraging Prophesee Event-Based Metavision sensing technologies.