[FEAT] Add U-2-Net

[BritishWerewolf]

Prerequisites

• [FEAT] Add bilinear resize to RawImage #1101
• [FEAT] Add Gaussian Blur to RawImage #1103
• [FEAT] Add dilate and erode operations to Tensor #1136

These PRs are required because the processors lean heavily on the new features.

---

Add support for the U-2-Net architecture which is a mask generation model; useful for background removal.
Both U-2-Net and ISNet can be used with these processors.

Below is an example of what the model will achieve when using the BritishWerewolf/U-2-Netp model.

Model	Original	Masked
U-2-Netp
IS-Net

If you would like to run this code, you can do with the BritishWerewolf/U-2-Net collection of models.

// Create the processor.
const processor = await AutoProcessor.from_pretrained(modelName);

// Create the model.
const model = await AutoModel.from_pretrained(modelName, {
	dtype: 'fp32',
});

// Process the image.
const image = await RawImage.read('https://huggingface.co/ybelkada/segment-anything/resolve/main/assets/car.png');
const processed_image = await processor(image);

// Run the model.
const output = await model({
	input: processed_image.pixel_values,
});

// Retrieve the mask, and scale it up to the size of the original image.
const mask = await RawImage.fromTensor(output.mask)
	.resize(image.width, image.height);

// Apply the mask to the image and save it.
const maskedImage = await image.putAlpha(mask);
maskedImage.save('masked_image.png');