Know-how for implementing edge AI that operates image × language AI with limited resources.
This article explains the implementation know-how for running small vLMs such as Moondream2, SmolVLM, and Qwen2-VL on Raspberry Pi 4/5 in edge operation. It covers llama.cpp + GGUF quantization, resolution and token control, memory optimization, and deployment in real projects. We will build image × language AI with limited resources. For more details, please refer to the technical column in the related links.
Inquire About This Product
basic information
【Technical Elements】 - Raspberry Pi 4/5 Edge AI - Small vLMs such as Moondream2 / SmolVLM / Qwen2-VL - llama.cpp + GGUF quantization - Resolution, token control, memory optimization 【Category】AI, Edge AI, Embedded *For more details, please refer to the technical column in the related links.
Price range
Delivery Time
Applications/Examples of results
【Intended Use】 - Image recognition and image captioning in offline environments - Utilization of image × language AI on edge devices - AI inference in low-resource environments 【Examples of Achievements】 - Implementation of a small vLM on Raspberry Pi and deployment in actual projects
Company information
Technosphere Co., Ltd. is a system development company based in Osaka that tackles customer challenges in advanced technology areas such as AI, IoT, and web system development. Since its founding in 2021, the company has leveraged a flat team structure to achieve a flexible and speedy development style. We are engaged in solutions that directly address social issues, such as image inspection AI and smart factory support systems.
![[Processing Example] Guide (2)](https://image.mono.ipros.com/public/product/image/bc0/2000278754/IPROS8852077070664670602.JPG?w=280&h=280)
