Gemma4-WebGPU / src /utils /coco-examples.ts
BryanBradfo's picture
Gemma 4 multimodal WebGPU detection Space
576d07a
Raw
History Blame Contribute Delete
1.19 kB
export interface CocoExample {
url: string;
prompt: string;
label: string;
}
export const DEFAULT_DETECTION_PROMPT =
'Detect all objects in this image. Return each detection as JSON: [{"box_2d": [y1, x1, y2, x2], "label": "object_name"}]';
const DETECT_PROMPT = DEFAULT_DETECTION_PROMPT;
const DESCRIBE_PROMPT =
"Describe this image in detail. What objects, people, and activities do you see?";
// Verified existing images in the Xenova/transformers.js-docs dataset (CORS enabled, stable)
const BASE = "https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main";
export const COCO_EXAMPLES: CocoExample[] = [
{
url: `${BASE}/cats.jpg`,
prompt: DETECT_PROMPT,
label: "Cats",
},
{
url: `${BASE}/city-streets.jpg`,
prompt: DETECT_PROMPT,
label: "Streets",
},
{
url: `${BASE}/corgi.jpg`,
prompt: DETECT_PROMPT,
label: "Corgi",
},
{
url: `${BASE}/football-match.jpg`,
prompt: DETECT_PROMPT,
label: "Football",
},
{
url: `${BASE}/airport.jpg`,
prompt: DETECT_PROMPT,
label: "Airport",
},
{
url: `${BASE}/butterfly.jpg`,
prompt: DESCRIBE_PROMPT,
label: "Butterfly",
},
];