Featuring 14 hours of recording on a 1-megapixel camera, available for download now
PARIS – December 7, 2020 – Prophesee SA, inventor of the most advanced neuromorphic vision systems, today announced the release of the most comprehensive Event-Based dataset to date. The dataset, which was spotlighted at the 2020 NeurIPS conference, consists of 14 hours of recordings on a 1-megapixel camera in automotive scenarios, plus 25M bounding boxes. It is available for immediate download here.
Event cameras encode visual information with high temporal precision, low data-rate, and high-dynamic range. Thanks to these characteristics, event cameras are particularly suited for scenarios with high motion, challenging lighting conditions and requiring low latency. However, due to the novelty of the field, the performance of event-based systems on many vision tasks is still lower compared to conventional frame-based solutions.
The main reasons for this performance gap are: the lower spatial resolution of event sensors, compared to frame cameras; the lack of large-scale training datasets; the absence of well-established deep learning architectures for event-based processing.
The Prophesee model, spotlighted at NeurIPS, outperforms by a large margin feed-forward event-based architectures. Moreover, its method does not require any reconstruction of intensity images from events, showing that training directly from raw events is possible, more efficient, and more accurate than passing through an intermediate intensity image.
Experiments on the algorithmically generated dataset introduced in this work, for which events and gray level images are available, show performance on par with that of highly tuned and studied frame-based detectors.
The dataset is split between train, test and val folders. Files consist of 60 seconds recordings that were cut from longer recording sessions. Cuts from a single recording session are all in the same training split. Each data file is a binary file in which events are encoded using 4 bytes (unsigned int32) for the timestamps and 4 bytes (unsigned int32) for the data, encoding is little-endian ordering. The data is composed of 14 bits for the x position, 14 bits for the y position and 1 bit for the polarity (encoded as -1/1).
The dataset is available with the machine learning module available in the company’s Metavision Intelligence Suite.
About Prophesee
Prophesee is the inventor of the world’s most advanced neuromorphic vision systems.
The company developed a breakthrough Event-Based Vision approach to machine vision. This new vision category allows for significant reductions of power, latency and data processing requirements to reveal what was invisible to traditional frame-based sensors until now.
Prophesee’s patented Metavision® sensors and algorithms mimic how the human eye and brain work to dramatically improve efficiency in areas such as autonomous vehicles, industrial automation, IoT, security and surveillance, and AR/VR.
Prophesee is based in Paris, with local offices in Grenoble, Shanghai, Tokyo and Silicon Valley. The company is driven by a team of more than 100 visionary engineers, holds more than 50 international patents and is backed by leading international investors including Sony, iBionext, 360 Capital Partners, Intel Capital, Robert Bosch Venture Capital, Supernova Invest, and European Investment Bank.
For more information visit: www.prophesee.ai.
