This is my first Hacktoberfest project. The week-one theme was "touch grass", so I built a small web app that checks whether you really went outside.
---How it works----
- Take or upload a photo.
- An open-source AI model (CLIP, running in your browser with Transformers.js) decides whether the photo is outdoors in nature.
- If it is, One get "Touch grass: confirmed"(loved writing that) . If not, you're sent back outside.
- The AI then names the plant from a list of 126 common plants, flowers and trees. Wikipedia supplies the origin and description.
- One can copy the result, download it as a .txt file, or save it as a PDF.
The photo never leaves the device. The model runs locally, and only the plant name is sent to Wikipedia.
-----What I learned----
- AI models can run directly in a web page, with no server or API key.
- The first visit downloads the model, so it's slow once, then faster.
- It isn't perfect. It was too strict about a close-up stock photo of a rose, but it passed a real photo of grass from my street. It only knows the plants on its list.
Next steps? more plants, a gallery of past nature notes, and a daily outdoor challenge. But a break is a must (my exams are over for now)
I built this with help from an AI assistant, and I tested it myself outdoors. If you try it, tell me what plant it found!
And as always, thank you for entertaining me, feedback appreciated.. Kbai
Top comments (0)