Campus Flora Finder
An offline-first local AI plant identification application built with Python and Streamlit. It uses Ollama to run local vision-language models (llava/moondream) for plant recognition, requiring no internet connection and preserving user privacy.
Timeline (last 90 days)
Campus Flora Finder is a plant identification tool submitted by developer shreshth2906 under the "Touch Grass" theme of the Hacktoberfest open-source AI challenge.
Campus Flora Finder runs entirely offline and does not send any personal data or photos to any cloud API.
Campus Flora Finder is built with Python, uses Streamlit for the frontend, and the core inference is provided by open-source-weight vision-language models (llava or moondream) running locally via Ollama.
The application's processing flow is: Streamlit receives an uploaded JPEG/PNG image and converts it to a byte stream, which is sent along with a botanical prompt to the local model endpoint for identification, returning structured taxonomic information within a few seconds.
When the Ollama daemon is not running or the model weights are not initialized, the application gives a clear prompt rather than crashing directly.
Locally run open-weight models allow the identification pipeline to work with no network connection at all, making it suitable for outdoor scenarios with weak signals.
Using open-weight local models avoids the token and image ingestion costs of cloud-based multimodal APIs, enabling zero-cost usage.
The pragmatic positioning of Campus Flora Finder should be as a science popularization entry point that sparks curiosity, rather than a professional identification tool whose accuracy can be trusted.
All Facts (8)
The pragmatic positioning of Campus Flora Finder should be as a science popularization entry point that sparks curiosity, rather than a professional identification tool whose accuracy can be trusted.
50%UnverifiedCampus Flora Finder runs entirely offline and does not send any personal data or photos to any cloud API.
50%UnverifiedCampus Flora Finder is built with Python, uses Streamlit for the frontend, and the core inference is provided by open-source-weight vision-language models (llava or moondream) running locally via Ollama.
50%UnverifiedThe application's processing flow is: Streamlit receives an uploaded JPEG/PNG image and converts it to a byte stream, which is sent along with a botanical prompt to the local model endpoint for identification, returning structured taxonomic information within a few seconds.
50%UnverifiedWhen the Ollama daemon is not running or the model weights are not initialized, the application gives a clear prompt rather than crashing directly.
50%UnverifiedLocally run open-weight models allow the identification pipeline to work with no network connection at all, making it suitable for outdoor scenarios with weak signals.
50%UnverifiedUsing open-weight local models avoids the token and image ingestion costs of cloud-based multimodal APIs, enabling zero-cost usage.
50%UnverifiedCampus Flora Finder is a plant identification tool submitted by developer shreshth2906 under the "Touch Grass" theme of the Hacktoberfest open-source AI challenge.
50%