Sometimes last month, Espressif Systems contacted me to send a mysterious new development kit together with a 10-year anniversary T-shirt, an offer I found hard to decline 🙂
So I got a new year present this morning delivered by DHL…
The letter lists the main specifications of ESP-EYE development board:
- WiSoC- ESP32 dual core Tensilica LX6 processor with WiFi and Bluetooth
- Memory – 8MB PSRAM
- Storage – 4MB flash
- Camera – 2MP OV2640 camera
- Audio – Microphone
- USB – 1x micro USB port for power and programming
- Misc – Reset, boot and function buttons, 2x LEDs
- Dimensions – 41 x 21 mm
It reminds me of ESP32-CAM camera board, except it adds a microphone, it’s even smaller, and as an official Espressif board, software support might be better. AI features such as face recognition or detection, and voice wake-up are supported out of the box.
The board ships with a micro USB cable, and a card with a block diagram, and a quick start guide explaining you need to press the side “function” button to collect face data.
One side of the board features the 2MP camera, reset and boot buttons, two LEDs, and the microphone (top to bottom),
While the other side reveals the chips used in the design: ESP32-D0WD, 8MB PSRAM, 4MB flash, CP2102 USB to TTL chip, as well as the on-board antenna, a u.FL connector for external antenna, the micro USB port, and function button.
That’s all I know for now, as I can’t find documentation on Espressif website yet, and ESP-EYE board is not for sale at this time. The boards and T-shirt just have been sent to various members of the community, and after testing they should become available.
Jean-Luc started CNX Software in 2010 as a part-time endeavor, before quitting his job as a software engineering manager, and starting to write daily news, and reviews full time later in 2011.
25 Replies to “ESP-EYE ESP32 AI Development Board Supports Face detection, Voice wakeup”
Micro SD card would have been nice.
Makes up for the last ESP32+audio board mentioned here a couple months ago, which got cancelled before it shipped.
how can to be members of the community??
In my case, it’s because I cover Espressif products often, but the most active people on the ESP forums and helping with software development likely received / will receive one. I suppose I may have been one of the first to have received it because I’m based in Asia.
How is this special ? I have had this board since August. How is this a prototype ?
You must be referring to the ESP32-CAM board, as I’ve never ever heard about ESP-EYE, and could not find any online references when I looked yesterday.
Any chance of getting a purchase link? Please?
Very cool! I took a look at the software for it and it uses OpenMV Cam code. Glad to see our efforts are enabling camera data processing on microcontrollers.
Since the ESP32 lacks an NPU accelerator, will Espressif launch a software-only SDK for AI? I wonder if this will be part of ESP-IDF or a new framework like the ESP-ADF for audio.
Wouldn’t know. I’d be…hesitant…to consider them more than a decent IP cam that MIGHT be able to do h.264 streaming.
This is more because of the Kendryte coming along being a much more muscular part capable of doing everything that the ESP32 can and then some for roughly similar pricing per quantity. Your AI isn’t going to be much there because there’s only so much compute power those Tensilica LX6’s at 200-ish MHz can do. It’s an impressive little part, but they’ve had the industry come along and one-up them again in only about 2-3 years time.
That being said, if they ship it, I’ll play with it.
The Kendryte K210 has no built-in wireless connectivity so I’m not sure about the comparison. The Sipeed M1w module adds an esp8285 for WiFi.
An ESP32 SDK for face recognition has been available for a few months:
i tested some of these older modules on our wiki
that’s great. thanx for posting.
do you know what’s the diff between ttgo t8 and ttgo zero (supposedly a new thing) btw?
Does it over heat like M5Stack cam module when streaming video for extended time?
The article says PSRAM is 4MBit. Where did you get the 8MB?
Which article? I got the 8MB in Espressif Systems letter in the first photo.
The linked page just below your numbers:
ESP32-CAM board specifications:
Wireless Module- ESP32-S WiFi 802.11 b/g/n + Bluetooth 4.2 LE module with PCB antenna, u.FL connector, 32Mbit SPI flash, 4MBit PSRAM
External Storage – micro SD card slot up to 4GB
ESP32-CAM and ESP-EYE are two different products, they just both happen to feature ESP32 and a camera.
How many faces can be enrolled in this?
Can I find examples showing how I may use this in my own way? I want the face recognition feature to trigger some other API end points (when a known face is found) to achieve different things. Any leads?
I’ve been following the getting started guide for ESP-EYE @ https://github.com/espressif/esp-who/blob/master/docs/en/get-started/ESP-EYE_Getting_Started_Guide.md
But none of the samples are working for me. Either I get no output from the camera, or errors to connect to the camera.
Did you manage to get it working?
can any one provide me link for software on which i will code for ESP-EYE
ESP-WHO. You can follow the getting started guide linked above.
If there was an OV2640 camera module that outputted h.264 encoded video – how hard to update the libraries on the esp32cam?? Yes I am trying to find one…maybe :).
That way (I am sure you know this) we can just pass the video stream through the esp32 very efficiently.
And if there was some sort of micro-gstreamer/webrtc library that could run on the esp32 (gstreamer runs on it now)- similar to the micro-RTSP library (that https://github.com/geeksville that Kevin Hester put together)… then some interesting things could happen – then you could do live streaming with the esp32 to most any device very easily.