yolov10b-doclaynet_ONNX_document-layout-analysis

ONNX weights for https://huggingface.co/hantian/yolo-doclaynet yolo10b

Usage (Transformers.js)

If you haven't already, you can install the Transformers.js JavaScript library from NPM using:

npm i @huggingface/transformers

Example: Perform object-detection with Oblix/yolov10b-doclaynet_ONNX_document-layout-analysis.

const model = await AutoModel.from_pretrained(
    "Oblix/yolov10b-doclaynet_ONNX_document-layout-analysis",
    {
        dtype: "fp32"
    }
);
const processor = await AutoProcessor.from_pretrained(
    "Oblix/yolov10b-doclaynet_ONNX_document-layout-analysis"
);
const url =
    "https://huggingface.co/DILHTWD/documentlayoutsegmentation_YOLOv8_ondoclaynet/resolve/main/sample1.png";
const image = await RawImage.read(url);
const { pixel_values, reshaped_input_sizes } = await processor(image);

// Run object detection
const { output0 } = await model({ images: pixel_values });
const predictions = output0.tolist()[0];

const threshold = 0.35;
const [newHeight, newWidth] = reshaped_input_sizes[0]; // Reshaped height and width
const [xs, ys] = [image.width / newWidth, image.height / newHeight]; // x and y resize scales
for (const [xmin, ymin, xmax, ymax, score, id] of predictions) {
    if (score < threshold) continue;

    // Convert to original image coordinates
    const bbox = [xmin * xs, ymin * ys, xmax * xs, ymax * ys]
        .map((x) => x.toFixed(2))
        .join(", ");
    console.log(
        // eslint-disable-next-line @typescript-eslint/no-explicit-any
        `Found "${(model.config as any).id2label[id]}" at [${bbox}] with score ${score.toFixed(
            2
        )}.`
    );
}

Result

Found "Text" at [53.75, 478.56, 623.46, 562.13] with score 0.98.
Found "Text" at [54.20, 593.64, 609.42, 637.15] with score 0.98.
Found "Text" at [53.98, 715.41, 621.06, 759.33] with score 0.98.
Found "Text" at [53.98, 247.44, 610.82, 277.49] with score 0.97.
Found "Title" at [53.64, 75.40, 551.96, 159.72] with score 0.97.
Found "List-item" at [55.56, 761.62, 607.48, 792.06] with score 0.97.
Found "List-item" at [56.05, 657.97, 614.57, 701.79] with score 0.97.
Found "Text" at [54.10, 195.40, 221.43, 211.88] with score 0.96.
Found "Text" at [54.25, 169.14, 95.17, 186.22] with score 0.95.
Found "Text" at [54.15, 222.11, 98.62, 237.74] with score 0.95.
Found "Text" at [53.73, 429.63, 412.82, 446.28] with score 0.95.
Found "Page-header" at [308.98, 10.07, 605.53, 34.59] with score 0.95.
Found "Section-header" at [54.18, 338.87, 102.68, 355.16] with score 0.95.
Found "List-item" at [55.75, 793.91, 519.29, 810.43] with score 0.95.
Found "Section-header" at [54.20, 453.01, 145.02, 469.42] with score 0.94.
Found "Text" at [56.76, 309.85, 316.43, 325.71] with score 0.93.
Found "List-item" at [55.62, 812.37, 445.03, 829.42] with score 0.92.
Found "Page-footer" at [308.43, 907.93, 374.03, 922.28] with score 0.92.
Found "Section-header" at [53.70, 567.21, 75.24, 584.85] with score 0.91.
Found "Text" at [56.26, 289.47, 415.46, 306.48] with score 0.80.
Found "Text" at [54.11, 365.35, 623.46, 407.97] with score 0.79.
Found "List-item" at [55.77, 638.84, 382.47, 655.46] with score 0.60.

Labels

• Caption [0]
• Footnote [1]
• Formula [2]
• List-item [3]
• Page-footer [4]
• Page-header [5]
• Picture [6]
• Section-header [7]
• Table [8]
• Text [9]
• Title [10]