SenseCAP Watcher has three big feature updates: self-hosted LLM with #llava and #llama models, custom local model training with @EdgeImpulse and finally push-to-talk functionality for task creation on the go.
Based on @EspressifSystems #esp32 S3 and #himax HX6538 (Cortex M55 + Ethos-U55), made by @SeeedStudioSZ Watcher runs a local neural network inference with #ai accelerator and if necessary sends the data for further analysis for a multi-modal #largelanguagemodel, making it an intelligent conduit for the LLMs!