TheYuriG•2mo ago

How can I do simple image recognition with Deno?

It's some very simple recognition, i don't need to recognize cats or dogs or people, just a few set of images within a bigger image and the quantity of each. Imagine counting the totals of a loot drop in game. It's that simple.

4 Replies

Mrcool 🇵🇸•2mo ago

You can explore hugging js, for example a simple object detection

import { HfInference } from "npm:@huggingface/inference";

const HF_TOKEN = ...
const inference = new HfInference(HF_TOKEN);
const result = await inference.objectDetection({
    data: await (await fetch(import.meta.resolve("./img.png"))).blob(),
});
console.log(result);

import { HfInference } from "npm:@huggingface/inference";

const HF_TOKEN = ...
const inference = new HfInference(HF_TOKEN);
const result = await inference.objectDetection({
    data: await (await fetch(import.meta.resolve("./img.png"))).blob(),
});
console.log(result);

and there are many models to choose from

TheYuriGOP•2mo ago

was afraid that using a LLM would be the only option, I've seen that option online

Mrcool 🇵🇸•2mo ago

There are simple ways like looking for pixels colors if a yellow object is inside a dark background you can find it

jeff.hykin•5w ago

I think this task is less simple than you might think. BTW hugging face does more than just LLMs