Smart mini solar camera
Battery-free solar cam with AI. © Xnor.ai.
Based in Seattle, start-up Xnor presented a prototype wireless mini camera that runs entirely on solar energy and can analyze the image captures. The sensor has a resolution of 320 x 320 pixels and is coupled with an FPGA integrated circuit to process images using artificial intelligence algorithms. The camera transmits data via LPWAN. With an optional battery, the device can store enough energy during the day to operate in low light conditions and at night. The device is not yet commercialized, but its unit manufacturing cost would be around 10 USD. Low cost and energy independence open the door to many new applications.
Traditional models of deep learning are computationally intensive. They customarily use GPUs and are often limited to running in the cloud. Xnor’s approach is to reconvert the models into a proprietary binary convolutional neural network called Xnor-Net. This allows models that are 10 times faster, 20 to 200 times more energy efficient and 8 to 15 times less memory-intensive than traditional deep learning models.
⇨ Circuit Breaker, “This wireless AI camera runs entirely on solar power.”