The WiFi AI camera development kit based on the Ingenic T31 offers a battery life of over a year.

 With products like the Eufy EverCam, WiFi security cameras have previously promised a year of battery life, but Innophase's Ingenic T31-based Smart Video Application Development Kit may allow for the development of even more power-efficient WiFi AI security cameras with AI processing that can last over a year.

The development kit combines the Ingenic T31 MIPS & RISC-V camera SoC with the Talaria TWO INP101x ultra-low-power (57A @ DTIM10) Wi-Fi & BLE wireless module, which is expected to last 14.4 months on a 3,000 mAh battery while operating at a 99.3% idle, 0.7 percent video capture ratio.

In addition to Linux drivers for the Ingenic T31, the package contains AI algorithms for person detection, facial identification/recognition, cry detection, vehicle detection, and pet detection, as well as fast start and user guidelines. The MIPS processor and Talaria TWO interacted via a "Dual-Stack" software implementation, which divides processing tasks between the two chips to keep the system's current consumption to a minimum. Device provisioning via BLE, AWS Cloud, and Azure Cloud connection, and support for firmware upgrades over the air are all included in the devkit (OTA).

• 1080p At 25fps, H.265 video capture
• Every day, there are six video capture events.
• Capture video for ten seconds
• Wi-Fi DTIM10, 3min Keep-Alive Active/Idle ratio – Wi-Fi idle connected: 99.93% per day; Daily Active Video Capture: 0.07 percent
• Breakdown of battery consumption: Wi-Fi idle connected: 30% per day, active video capture: 70% per day
• Battery capacity: 3,000mAh @ 3.6V

ADK Core Components

Ingenic T31 Video Processor

• CPU: Ultra high frequency, up to 1.5GHz, Vector Deep Learning accelerator based on SIMD128, RiscV independent lite core
• Video Encoder: H.264/H.265/MJPEG encoder, Maximum 2592*1920@30fps
• Starlight ISP: Dedicated optimizations for low light and surveillance scenarios, Upgraded 2D / 3D noise reduction, Distortion correction
• AI Algorithm: Supports deep learning algorithms with high precision and good flexibility - Human detection, Facial detection/recognition, Cry detection, Vehicle detection, Pet detection
• Memory: Capacity of 512Mbit or 1Gbit
• Audio: Integrated Audio Codec, Support I2S Interface, Echo cancellation

 InnoPhase Talaria TWO Wi-Fi+BLE

• Wi-Fi: 802.11 b/g/n, up to MCS7 Single-stream (1x1)
• BLE: BLE 5.0 w/ Advanced Features: 2Mbps PHY, LE Coding (Long-Range), Extended Advertising
• Frequency Supported: 2.4GHz
• Application Processor: Arm Cortex-M3, 80MHz
• Embedded Memories: 512KB SRAM, 2MB Flash
• Peripheral Interface to T31: SDIO
• Power Consumption: Ultra-low-power Wi-Fi to maintain internet/cloud connectivity for
  battery-based products
• Ultra-Low Power Wi-Fi: 57uA (DTIM10), 31mA (Rx, 802.11b), 81mA (Tx, 802.11n)
• Optimized Software: exclusive “Dual-Stack” SW implementation shares processing responsibilities between Talaria TWO and T31 to reduce system current consumption to the lowest level

Under those circumstances, the kit should last 1.16 years or 14.4 months on a single charge. WiFi idle power consumption is very low, thus active video capture consumes the most power. This situation amounts to just 6 captures of 10 seconds every day, or a one-minute movie, which is unlikely in most circumstances. But, in most cases, it should be easy to develop AI cameras that last several months on a charge.

The kit is restricted in number, and access to the InnoPhase SDK involves signing a non-disclosure agreement (NDA). The product website includes further information on the development kit.