Run model on ESP32-CAM

[lynkos]

• Deploy model #47
• Increase model accuracy #41
• Minimize size of model weights #46
  • Size varies depending on format
  • Ideal format TFLite Edge TPU?

Advantages of TFLite Edge TPU include:

• Optimized Performance on Edge Devices
  • Achieves high-speed neural networking performance through quantization, model optimization, hardware acceleration, and compiler optimization
  • Minimalistic architecture contributes to its smaller size and cost-efficiency
• High Computational Throughput
  • Combines specialized hardware acceleration and efficient runtime execution to achieve high computational throughput
  • Well-suited for deploying ML models with stringent performance requirements on edge devices
• Efficient Matrix Computations
  • Optimized for matrix operations (crucial for neural network computations)
  • This efficiency is key in ML models, particularly those requiring numerous and complex matrix multiplications and transformations
• Deployment
  • On-Device: Directly deploy on mobile and embedded devices, which allows the models to execute directly on the hardware (eliminating the need for cloud connectivity)
  • Edge Computing with Cloud TensorFlow TPUs: Offload inference tasks to cloud servers equipped with TPUs for scenarios where edge devices have limited processing capabilities
  • Hybrid: Versatile and scalable solution for deploying ML models; includes on-device processing for quick responses and cloud deployment/computing for more complex computations