transformers.js/examples/segment-anything-client/worker.js

112 lines
3.3 KiB
JavaScript

import { SamModel, AutoProcessor, RawImage, Tensor } from '@xenova/transformers';
// We adopt the singleton pattern to enable lazy-loading of the model and processor.
export class SegmentAnythingSingleton {
static model_id = 'Xenova/slimsam-77-uniform';
static model;
static processor;
static getInstance() {
if (!this.model) {
this.model = SamModel.from_pretrained(this.model_id, {
dtype: {
vision_encoder: 'fp16',
prompt_encoder_mask_decoder: 'q8',
},
device: {
vision_encoder: 'webgpu',
prompt_encoder_mask_decoder: 'wasm',
}
});
}
if (!this.processor) {
this.processor = AutoProcessor.from_pretrained(this.model_id);
}
return Promise.all([this.model, this.processor]);
}
}
// State variables
let imageEmbeddings = null;
let imageInputs = null;
let ready = false;
self.onmessage = async (e) => {
const [model, processor] = await SegmentAnythingSingleton.getInstance();
if (!ready) {
// Indicate that we are ready to accept requests
ready = true;
self.postMessage({
type: 'ready',
});
}
const { type, data } = e.data;
if (type === 'reset') {
imageInputs = null;
imageEmbeddings = null;
} else if (type === 'segment') {
// Indicate that we are starting to segment the image
self.postMessage({
type: 'segment_result',
data: 'start',
});
// Read the image and recompute image embeddings
const image = await RawImage.read(e.data.data);
imageInputs = await processor(image);
imageEmbeddings = await model.get_image_embeddings(imageInputs)
// Indicate that we have computed the image embeddings, and we are ready to accept decoding requests
self.postMessage({
type: 'segment_result',
data: 'done',
});
} else if (type === 'decode') {
// Prepare inputs for decoding
const reshaped = imageInputs.reshaped_input_sizes[0];
const points = data.map(x => [x.point[0] * reshaped[1], x.point[1] * reshaped[0]])
const labels = data.map(x => BigInt(x.label));
const input_points = new Tensor(
'float32',
points.flat(Infinity),
[1, 1, points.length, 2],
)
const input_labels = new Tensor(
'int64',
labels.flat(Infinity),
[1, 1, labels.length],
)
// Generate the mask
const { pred_masks, iou_scores } = await model({
...imageEmbeddings,
input_points,
input_labels,
})
// Post-process the mask
const masks = await processor.post_process_masks(
pred_masks,
imageInputs.original_sizes,
imageInputs.reshaped_input_sizes,
);
// Send the result back to the main thread
self.postMessage({
type: 'decode_result',
data: {
mask: RawImage.fromTensor(masks[0][0]),
scores: iou_scores.data,
},
});
} else {
throw new Error(`Unknown message type: ${type}`);
}
}