feat(web): show ocr text boxes in panoramas (#25727)

This commit is contained in:
Mees Frensel
2026-02-19 02:04:18 +01:00
committed by GitHub
parent 1f8359ead4
commit 84f7fb63ee
4 changed files with 170 additions and 83 deletions
@@ -425,7 +425,6 @@
const showOcrButton = $derived(
$slideshowState === SlideshowState.None &&
asset.type === AssetTypeEnum.Image &&
!(asset.exifInfo?.projectionType === 'EQUIRECTANGULAR') &&
!assetViewerManager.isShowEditor &&
ocrManager.hasOcrData,
);
@@ -1,6 +1,6 @@
<script lang="ts">
import type { OcrBox } from '$lib/utils/ocr-utils';
import { calculateBoundingBoxDimensions } from '$lib/utils/ocr-utils';
import { calculateBoundingBoxMatrix } from '$lib/utils/ocr-utils';
type Props = {
ocrBox: OcrBox;
@@ -8,28 +8,19 @@
let { ocrBox }: Props = $props();
const dimensions = $derived(calculateBoundingBoxDimensions(ocrBox.points));
const dimensions = $derived(calculateBoundingBoxMatrix(ocrBox.points));
const transform = $derived(
`translate(${dimensions.minX}px, ${dimensions.minY}px) rotate(${dimensions.rotation}deg) skew(${dimensions.skewX}deg, ${dimensions.skewY}deg)`,
);
const transformOrigin = $derived(
`${dimensions.centerX - dimensions.minX}px ${dimensions.centerY - dimensions.minY}px`,
const transform = $derived(`matrix3d(${dimensions.matrix.join(',')})`);
// Fits almost all strings within the box, depends on font family
const fontSize = $derived(
`max(var(--text-sm), min(var(--text-6xl), ${(1.4 * dimensions.width) / ocrBox.text.length}px))`,
);
</script>
<div class="absolute group left-0 top-0 pointer-events-none">
<!-- Bounding box with CSS transforms -->
<div class="absolute left-0 top-0">
<div
class="absolute border-2 border-blue-500 bg-blue-500/10 cursor-pointer pointer-events-auto transition-all group-hover:bg-blue-500/30 group-hover:border-blue-600 group-hover:border-[3px]"
style="width: {dimensions.width}px; height: {dimensions.height}px; transform: {transform}; transform-origin: {transformOrigin};"
></div>
<!-- Text overlay - always rendered but invisible, allows text selection and copy -->
<div
class="absolute flex items-center justify-center text-transparent text-sm px-2 py-1 pointer-events-auto cursor-text whitespace-pre-wrap wrap-break-word select-text group-hover:text-white group-hover:bg-black/75 group-hover:z-10"
style="width: {dimensions.width}px; height: {dimensions.height}px; transform: {transform}; transform-origin: {transformOrigin};"
class="absolute flex items-center justify-center text-transparent text-sm border-2 border-blue-500 bg-blue-500/10 px-2 py-1 pointer-events-auto cursor-text whitespace-pre-wrap wrap-break-word select-text transition-all hover:text-white hover:bg-black/60 hover:border-blue-600 hover:border-3"
style="font-size: {fontSize}; width: {dimensions.width}px; height: {dimensions.height}px; transform: {transform}; transform-origin: 0 0;"
>
{ocrBox.text}
</div>
@@ -2,8 +2,10 @@
import { shortcuts } from '$lib/actions/shortcut';
import AssetViewerEvents from '$lib/components/AssetViewerEvents.svelte';
import { assetViewerManager } from '$lib/managers/asset-viewer-manager.svelte';
import { ocrManager, type OcrBoundingBox } from '$lib/stores/ocr.svelte';
import { boundingBoxesArray, type Faces } from '$lib/stores/people.store';
import { alwaysLoadOriginalFile } from '$lib/stores/preferences.store';
import { calculateBoundingBoxMatrix, getOcrBoundingBoxesAtSize, type Point } from '$lib/utils/ocr-utils';
import {
EquirectangularAdapter,
Viewer,
@@ -27,6 +29,17 @@
strokeLinejoin: 'round',
};
// Adapted as well as possible from classlist 'border-2 border-blue-500 bg-blue-500/10 hover:border-blue-600 hover:border-3'
const OCR_BOX_SVG_STYLE = {
fill: 'var(--color-blue-500)',
fillOpacity: '0.1',
stroke: 'var(--color-blue-500)',
strokeWidth: '2px',
};
const OCR_TOOLTIP_HTML_CLASS =
'flex items-center justify-center text-white bg-black/50 cursor-text pointer-events-auto whitespace-pre-wrap wrap-break-word select-text';
type Props = {
panorama: string | { source: string };
originalPanorama?: string | { source: string };
@@ -96,6 +109,59 @@
}
});
$effect(() => {
updateOcrBoxes(ocrManager.showOverlay, ocrManager.data);
});
/** Use updateOnly=true on zoom, pan, or resize. */
const updateOcrBoxes = (showOverlay: boolean, ocrData: OcrBoundingBox[], updateOnly = false) => {
if (!viewer || !viewer.state.textureData || !viewer.getPlugin(MarkersPlugin)) {
return;
}
const markersPlugin = viewer.getPlugin<MarkersPlugin>(MarkersPlugin);
if (!showOverlay) {
markersPlugin.clearMarkers();
return;
}
if (!updateOnly) {
markersPlugin.clearMarkers();
}
const boxes = getOcrBoundingBoxesAtSize(ocrData, {
width: viewer.state.textureData.panoData.croppedWidth,
height: viewer.state.textureData.panoData.croppedHeight,
});
for (const [index, box] of boxes.entries()) {
const points = box.points.map((p) => texturePointToViewerPoint(viewer, p));
const { matrix, width, height } = calculateBoundingBoxMatrix(points);
const fontSize = (1.4 * width) / box.text.length; // fits almost all strings within the box, depends on font family
const transform = `matrix3d(${matrix.join(',')})`;
const content = `<div class="${OCR_TOOLTIP_HTML_CLASS}" style="font-size: ${fontSize}px; width: ${width}px; height: ${height}px; transform: ${transform}; transform-origin: 0 0;">${box.text}</div>`;
if (updateOnly) {
markersPlugin.updateMarker({
id: `box_${index}`,
polygonPixels: box.points.map((b) => [b.x, b.y]),
tooltip: { content },
});
} else {
markersPlugin.addMarker({
id: `box_${index}`,
polygonPixels: box.points.map((b) => [b.x, b.y]),
svgStyle: OCR_BOX_SVG_STYLE,
tooltip: { content, trigger: 'click' },
});
}
}
};
const texturePointToViewerPoint = (viewer: Viewer, point: Point) => {
const spherical = viewer.dataHelper.textureCoordsToSphericalCoords({ textureX: point.x, textureY: point.y });
return viewer.dataHelper.sphericalCoordsToViewerCoords(spherical);
};
const onZoom = () => {
viewer?.animate({ zoom: assetViewerManager.zoom > 1 ? 50 : 83.3, speed: 250 });
};
@@ -160,7 +226,20 @@
viewer.addEventListener(events.ZoomUpdatedEvent.type, zoomHandler, { passive: true });
}
return () => viewer.removeEventListener(events.ZoomUpdatedEvent.type, zoomHandler);
const onReadyHandler = () => updateOcrBoxes(ocrManager.showOverlay, ocrManager.data, false);
const updateHandler = () => updateOcrBoxes(ocrManager.showOverlay, ocrManager.data, true);
viewer.addEventListener(events.ReadyEvent.type, onReadyHandler);
viewer.addEventListener(events.PositionUpdatedEvent.type, updateHandler);
viewer.addEventListener(events.SizeUpdatedEvent.type, updateHandler);
viewer.addEventListener(events.ZoomUpdatedEvent.type, updateHandler, { passive: true });
return () => {
viewer.removeEventListener(events.ReadyEvent.type, onReadyHandler);
viewer.removeEventListener(events.PositionUpdatedEvent.type, updateHandler);
viewer.removeEventListener(events.SizeUpdatedEvent.type, updateHandler);
viewer.removeEventListener(events.ZoomUpdatedEvent.type, updateHandler);
viewer.removeEventListener(events.ZoomUpdatedEvent.type, zoomHandler);
};
});
onDestroy(() => {
@@ -176,3 +255,25 @@
<svelte:document use:shortcuts={[{ shortcut: { key: 'z' }, onShortcut: onZoom, preventDefault: true }]} />
<div class="h-full w-full mb-0" bind:this={container}></div>
<style>
/* Reset the default tooltip styling */
:global(.psv-tooltip) {
top: 0 !important;
left: 0 !important;
background: none;
box-shadow: none;
width: 0;
height: 0;
}
:global(.psv-tooltip-content) {
font: var(--font-normal);
padding: 0;
text-shadow: none;
}
:global(.psv-tooltip-arrow) {
display: none;
}
</style>