Sometimes new boards pop up on the radar almost entirely out of the blue, like the new Espressif ESP-EYE has done. This one made its way into the hands of Jean-Luc Aufranc, from the CNX Software blog, over the Christmas vacation.
“…this is Espressif’s new AI development board featuring voice wakeup, as well as face detection and recognition. It has been built around our flagship chip, ESP32, and it is also equipped with a 2-Megapixel OV 2640 Camera, Microphone, 4 MB Flash, 8MB PSRAM, Micro USB and LED lights on a 21mm × 41mm board.”
On the face of it, this is an intriguing addition to Espressif’s product lineup. With most of the major machine learning tools being heavily optimised for Arm-based architectures, Espressif with its Tensilica-based chips is in danger of being left behind when it comes to machine learning on the edge. One of the reasons I was so excited by Seeed Studio’s Air602 module released late last year is that it’s a price comparable ESP8266 competitor that’s built on Arm.
On the surface, the board itself looks a lot like a neater, more finished version of the prototype Ai-Thinker CAM board we saw towards the middle of last year, as well a few of the other ESP32-based camera boards I’ve seen. What’s different here is of course that this is, presumably, going to be an “official” Espressif product and that implies a better level of software support than we’ve seen before. But, at least right now, there doesn’t seem to be any details of the software stack they’re using on the board, although it’s likely to be their own ESP-WHO framework. Which is sort of interesting.
In fact, there really doesn’t seem to be many details of the new board at all; there certainly isn’t any information available on the Espressif website. Beyond the handful of boards that got sent out to the community over the Christmas break, the new board hasn’t made it into more general circulation. It isn’t yet for sale, and there is no word on pricing or even a release date.
So, now all I have to do is figure out how to get my hands on one…?
[h/t: CNX Software]