How far are event-based (neuromorphic) vision sensors from mainstream use in phones, cars, robots and factories?

The road so far

7 of 10 steps done, 2 in progress.

1. Done · Sept 2021Event sensor announced for industrial use

   Sony and Prophesee announced the IMX636 (0.92 megapixel) and IMX637 (0.33 megapixel) event sensors in September 2021, with a 4.86 micrometre pixel. Sony said they were designed for industrial equipment. Sony's announcement gave no performance figures beyond resolution and pixel size. [4]

2. DoneDeveloper kit for the HD event sensor

   Prophesee's EVK4 kit uses the IMX636 with 1280 by 720 pixels and pixel latency below 100 microseconds. Dynamic range is above 86 dB measured from 5 to 100,000 lux, or above 120 dB when measured from the 0.08 lux low-light cutoff. Prophesee does not state the sensor's power consumption or peak event rate on the page, and the page gives no launch date. [5]

3. Done · Oct 2020Samsung event sensor reported in a research paper

   Samsung reported a 1280 by 960 dynamic vision sensor with 4.95 micrometre pixels in a 2020 conference paper. The authors report 122 nW per pixel, 1.25 times smaller than the previous state of the art, as the Image Sensors World post quotes the paper. The post gives no baseline figure. No Samsung event sensor product was found in the sources reviewed. [6]

4. Done · Mar 2025Industrial event cameras on sale

   Prophesee lists IDS's uEye EVS industrial camera series, launched on 5 March 2025 with the IMX636, and a CenturyArks camera module launched on 28 October 2025. In March 2026 IDS and Prophesee signed a letter of intent to develop next-generation industrial vision systems, with no products, volumes or dates named. [7][8]

5. In progress · Apr 2021Event sensor in car driver monitoring

   In April 2021 DTS, an Xperi subsidiary, announced its AutoSense driver and occupant monitoring system, built on neuromorphic (event-based) vision. The release named no automaker or vehicle model, and the sources reviewed do not confirm a production car using it. In February 2021 Terranet's VoxelFlow system, built from three event cameras and a continuous laser scanner, was presented jointly with Mercedes-Benz at STARTUP AUTOBAHN. The post does not say Mercedes-Benz deploys the system. The figures are the company's, and no production vehicle is confirmed. [10][11]

6. Done · May 2024Event and frame cameras combined for car detection

   Gehrig and Scaramuzza pair a 20 fps RGB camera with an event camera. The authors report that the pair matches the latency of a 5,000 fps camera while using the bandwidth of a 45 fps camera, and they propose a hybrid event and frame object detector. The full text was not reachable, so these figures rest on the abstract as reported in a search summary. The sources reviewed do not show a real-time test on a moving vehicle. [1]

7. Done · Mar 2020Drone avoids fast obstacles using an event camera

   Falanga, Kleber and Scaramuzza report a quadrotor that used event cameras to dodge approaching obstacles. The authors give an overall latency of about 3.5 ms for the algorithm. The test used a research quadrotor, not a commercial product. [2]

8. Done · Mar 2023Drone flight controlled by a spiking network

   Paredes-Vallés and colleagues report a spiking neural network trained on raw event-camera data that runs on board Intel's Loihi neuromorphic processor at an execution frequency of 200 Hz. The 2023 preprint reports 27 microjoules per inference. A 2024 journal version was not opened in this review. [3]

9. Not startedReal-time test of event and frame fusion on a vehicle

   The hybrid automotive study did not report a real-time system tested on a moving vehicle, as far as the sources reviewed show. No on-vehicle trial with published end-to-end latency has been found. [1]

10. In progressShared software and benchmarks for event data

    Surveys describe many methods for event data, including deep learning. The sources reviewed do not show a single representation or benchmark adopted across vendors. Zheng and colleagues state that deep learning for event-based vision still lacks a taxonomy of techniques. [12][13]

Why it matters

Event sensors respond in microseconds and handle very bright and very dark areas in the same scene. If they become cheap and easy to program, machines could react faster on small batteries and send less data. Most vision software and training data still assume frames, so the software and data gaps will decide how quickly this moves beyond specialist use.

What stands in the way

• Most vision software and training data assume full frames. Event streams need new representations and algorithms, and the surveys say these are still being developed.
• Labelled event data and shared benchmarks. The sources reviewed do not show a common labelled dataset for automotive event detection. The full text of the 2024 hybrid automotive study was not reachable, so how its training labels were made could not be checked.
• Size, cost and volume. In 2019 Prophesee's Verre said the packaged sensor was too big for mainstream automotive use, and that cost and volume had to improve before mass markets could be reached.
• Vendor figures often lack test conditions. Prophesee's sensor page gives no power figure, and throughput and power claims from vendors have not been tested independently in the sources reviewed.
• Validation in cars. The strongest automotive result has not been shown on a moving vehicle in the sources reviewed. No real-time on-vehicle test with published latency has been found.
• Controlled settings. iniVation's DVXplorer documentation says the camera is intended for laboratory and industrial use under controlled conditions.

Published forecasts

What others have predicted. These are their estimates, not ours.

Signals to watch

• A named automaker or Tier 1 supplier announcing a production vehicle model that uses an event sensor.
• On-vehicle, real-time tests of event and frame fusion that report end-to-end latency and power.
• An event sensor in a phone or consumer device, with the sensor maker named.
• A common data format and public benchmark for event-based detection adopted by several vendors.
• The IDS and Prophesee partnership naming products, volumes or dates, as the 2026 letter of intent aims to develop next-generation industrial systems.

Sources

1. peer reviewed paper
   Low-latency automotive vision with event cameras
   Daniel Gehrig, Davide Scaramuzza. Nature, 29 May 2024.
2. peer reviewed paper
   Dynamic obstacle avoidance for quadrotors with event cameras
   Davide Falanga, Kevin Kleber, Davide Scaramuzza. Science Robotics, Mar 2020.
3. preprint
   Fully neuromorphic vision and control for autonomous drone flight
   Federico Paredes-Vallés, Jesse Hagenaars, Julien Dupeyroux et al.. arXiv (preprint), 15 Mar 2023.
4. news
   Sony Announces Two 4.86um Pixel Event-Based Sensors Developed with Prophesee
   Image Sensors World (blog), 9 Sept 2021.
5. company announcement
   Event Camera Evaluation Kit 4 HD IMX636 Prophesee-Sony
   Prophesee, Apr 2022.
6. news
   Samsung Presents 1.3MP Event-Driven Sensor with 4.95um Pixels
   Yunjae Suh. Image Sensors World (blog), quoting a 2020 conference paper, 3 Oct 2020.
7. company announcement
   Everything That Happened at Prophesee in 2025
   Prophesee, 7 Jan 2026.
8. company announcement
   IDS and Prophesee to Advance Event-Based Industrial Vision
   Prophesee, 11 Mar 2026.
9. news
   Event-Driven Vision Hits Production Lines
   EE Times, 25 Sept 2019.
10. company announcement
    Xperi Develops World-first In-cabin Monitoring Technologies on Neuromorphic Vision Systems
    Prophesee (copy of the DTS and Xperi release), 21 Apr 2021.
11. news
    Terranet and Mercedes-Benz Present Quick Reacting Event-Based Camera System
    Image Sensors World (blog), reposting a Terranet release of 25 February 2021, 16 Mar 2021.
12. review article
    Event-based Vision: A Survey
    Guillermo Gallego, Tobi Delbruck, Garrick Orchard et al.. IEEE Transactions on Pattern Analysis and Machine Intelligence, Jan 2022.
13. preprint
    Deep Learning for Event-based Vision: A Comprehensive Survey and Benchmarks
    Xu Zheng, Yexin Liu, Yunfan Lu et al.. arXiv, 17 Feb 2023.
14. other
    DVXplorer product documentation
    iniVation AG, 26 May 2026.

Last reviewed 21 Oct 2026.