boc/quixzoom-video-pipeline/dataset-builder/builder.js

/**
 * QUIXZOOM Video Pipeline — Dataset Builder
 *
 * Bygger träningsdataset från AI-annoterade frames:
 * - Roads
 * - Sidewalks
 * - Crosswalks
 * - Buildings
 * - Trees
 * - Utility Infrastructure
 * - Traffic Infrastructure
 * - Signage
 * - Lighting
 * - Accessibility
 * - Commercial Areas
 * - Urban Density
 *
 * Output: JSONL-format för ML-träning
 */

const fs = require('fs').promises;
const path = require('path');

/**
 * Dataset-konfiguration
 */
const DATASET_CONFIG = {
  version: '1.0.0',
  format: 'jsonl',
  splits: {
    train: 0.7,
    validation: 0.15,
    test: 0.15,
  },
};

/**
 * Bygg dataset från annoterade frames
 */
async function buildDataset(inputDir, outputDir, options = {}) {
  console.log(`[DATASET] Building dataset from: ${inputDir}`);

  // Läs alla annotations
  const annotations = await loadAnnotations(inputDir);
  console.log(`[DATASET] Loaded ${annotations.length} annotations`);

  // Filtrera efter kvalitet
  const qualityFilter = options.minQuality || 60;
  const filtered = annotations.filter(a =>
    (a.aiAnalysis?.qualityScore?.overall || 0) >= qualityFilter
  );
  console.log(`[DATASET] ${filtered.length} annotations passed quality filter (${qualityFilter}+)`);

  // Bygg dataset för varje kategori
  const datasets = {
    roads: buildRoadDataset(filtered),
    sidewalks: buildSidewalkDataset(filtered),
    crosswalks: buildCrosswalkDataset(filtered),
    buildings: buildBuildingDataset(filtered),
    trees: buildTreeDataset(filtered),
    utilityInfrastructure: buildUtilityDataset(filtered),
    trafficInfrastructure: buildTrafficDataset(filtered),
    signage: buildSignageDataset(filtered),
    lighting: buildLightingDataset(filtered),
    accessibility: buildAccessibilityDataset(filtered),
    commercialAreas: buildCommercialDataset(filtered),
    urbanDensity: buildDensityDataset(filtered),
  };

  // Spara dataset
  await saveDatasets(datasets, outputDir);

  // Generera statistik
  const stats = generateStats(datasets);
  await fs.writeFile(
    path.join(outputDir, 'stats.json'),
    JSON.stringify(stats, null, 2)
  );

  console.log(`[DATASET] Complete. Saved to: ${outputDir}`);

  return { datasets, stats };
}

/**
 * Läs annotations från katalog
 */
async function loadAnnotations(inputDir) {
  const files = await fs.readdir(inputDir);
  const jsonFiles = files.filter(f => f.endsWith('.json'));

  const annotations = [];
  for (const file of jsonFiles) {
    const content = await fs.readFile(path.join(inputDir, file), 'utf8');
    annotations.push(JSON.parse(content));
  }

  return annotations;
}

/**
 * Bygg väg-dataset
 */
function buildRoadDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const roadSurface = annotation.roadSurface;
    const segmentation = annotation.segmentation;

    if (!roadSurface && !segmentation) continue;

    const sample = {
      image: annotation.framePath,
      labels: {
        surface_type: roadSurface?.topSurface?.type || 'unknown',
        surface_confidence: roadSurface?.topSurface?.confidence || 0,
        has_cracks: roadSurface?.hasCracks || false,
        crack_severity: annotation.cracks?.severity || 'none',
        road_pixels: segmentation?.distribution?.road || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg trottoar-dataset
 */
function buildSidewalkDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const sidewalk = annotation.sidewalk;
    const segmentation = annotation.segmentation;

    if (!sidewalk && !segmentation) continue;

    const sample = {
      image: annotation.framePath,
      labels: {
        present: sidewalk?.present || false,
        width: sidewalk?.width || 0,
        condition: sidewalk?.condition || 'unknown',
        wheelchair_accessible: sidewalk?.accessibility?.wheelchair || false,
        tactile_paving: sidewalk?.accessibility?.tactile_paving || false,
        sidewalk_pixels: segmentation?.distribution?.sidewalk || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg övergångsställe-dataset
 */
function buildCrosswalkDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const objects = annotation.objects?.objects || [];
    const trafficSigns = annotation.trafficSigns || [];

    const hasCrosswalk = objects.some(o => o.class === 'pedestrian crossing') ||
                        trafficSigns.some(s => s.type === 'pedestrian_crossing');

    if (!hasCrosswalk) continue;

    const sample = {
      image: annotation.framePath,
      labels: {
        present: true,
        type: 'zebra', // Kan detekteras mer specifikt
        traffic_light_nearby: objects.some(o => o.class === 'traffic light'),
        sign_present: trafficSigns.some(s => s.type === 'pedestrian_crossing'),
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg byggnads-dataset
 */
function buildBuildingDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const buildings = annotation.buildings;
    const segmentation = annotation.segmentation;
    const ocr = annotation.ocr;

    const sample = {
      image: annotation.framePath,
      labels: {
        count: buildings?.count || 0,
        types: buildings?.buildings?.map(b => b.type) || [],
        heights: buildings?.buildings?.map(b => b.height) || [],
        building_pixels: segmentation?.distribution?.building || 0,
        has_hotel: ocr?.storefronts?.some(s => /HOTEL/i.test(s)) || false,
        has_restaurant: ocr?.storefronts?.some(s => /RESTAURANT/i.test(s)) || false,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg träd-dataset
 */
function buildTreeDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const vegetation = annotation.vegetation;
    const segmentation = annotation.segmentation;

    const sample = {
      image: annotation.framePath,
      labels: {
        tree_count: vegetation?.treeCount || 0,
        coverage: vegetation?.coverage || 0,
        health: vegetation?.health || 'unknown',
        vegetation_pixels: segmentation?.distribution?.vegetation || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg utility-infrastruktur-dataset
 */
function buildUtilityDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const poles = annotation.poles || [];
    const utilityBoxes = annotation.utilityBoxes || [];

    const sample = {
      image: annotation.framePath,
      labels: {
        pole_count: poles.length,
        pole_types: poles.map(p => p.type),
        utility_box_count: utilityBoxes.length,
        utility_box_types: utilityBoxes.map(b => b.type),
        has_overhead_wires: poles.some(p => p.type === 'utility_pole'),
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg trafik-infrastruktur-dataset
 */
function buildTrafficDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const objects = annotation.objects?.objects || [];
    const trafficSigns = annotation.trafficSigns || [];

    const sample = {
      image: annotation.framePath,
      labels: {
        traffic_light_count: objects.filter(o => o.class === 'traffic light').length,
        traffic_sign_count: trafficSigns.length,
        sign_types: trafficSigns.map(s => s.type),
        has_stop_sign: trafficSigns.some(s => s.type === 'stop'),
        has_speed_limit: trafficSigns.some(s => s.type === 'speed_limit'),
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg skylt-dataset
 */
function buildSignageDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const ocr = annotation.ocr;
    const trafficSigns = annotation.trafficSigns;

    const sample = {
      image: annotation.framePath,
      labels: {
        text_detected: ocr?.text || '',
        text_confidence: ocr?.confidence || 0,
        signs: ocr?.signs || [],
        storefronts: ocr?.storefronts || [],
        traffic_signs: trafficSigns?.map(s => s.type) || [],
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg belysnings-dataset
 */
function buildLightingDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const lighting = annotation.lighting;

    const sample = {
      image: annotation.framePath,
      labels: {
        street_light_count: lighting?.streetLights?.length || 0,
        building_light_count: lighting?.buildingLights?.length || 0,
        time_of_day: lighting?.timeOfDay || 'unknown',
        natural_light_level: lighting?.naturalLight || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg tillgänglighets-dataset
 */
function buildAccessibilityDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const accessibility = annotation.accessibility;
    const sidewalk = annotation.sidewalk;

    const sample = {
      image: annotation.framePath,
      labels: {
        wheelchair_ramp: accessibility?.wheelchairRamp || false,
        tactile_paving: accessibility?.tactilePaving || false,
        audible_signals: accessibility?.audibleSignals || false,
        braille_signage: accessibility?.brailleSignage || false,
        accessible_parking: accessibility?.accessibleParking || false,
        overall_score: accessibility?.overallScore || 0,
        sidewalk_width: sidewalk?.width || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg kommersiellt område-dataset
 */
function buildCommercialDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const storefronts = annotation.storefronts || [];
    const ocr = annotation.ocr;

    const sample = {
      image: annotation.framePath,
      labels: {
        business_count: storefronts.length,
        business_types: storefronts.map(s => s.type),
        business_names: storefronts.map(s => s.name),
        open_businesses: storefronts.filter(s => s.open).length,
        has_hotel: ocr?.storefronts?.some(s => /HOTEL/i.test(s)) || false,
        has_restaurant: ocr?.storefronts?.some(s => /RESTAURANT/i.test(s)) || false,
        has_cafe: ocr?.storefronts?.some(s => /CAFE/i.test(s)) || false,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Bygg urban densitet-dataset
 */
function buildDensityDataset(annotations) {
  const samples = [];

  for (const annotation of annotations) {
    const objects = annotation.objects;
    const segmentation = annotation.segmentation;

    const sample = {
      image: annotation.framePath,
      labels: {
        person_count: objects?.grouped?.person?.length || 0,
        vehicle_count: (
          (objects?.grouped?.car?.length || 0) +
          (objects?.grouped?.motorcycle?.length || 0) +
          (objects?.grouped?.bus?.length || 0) +
          (objects?.grouped?.truck?.length || 0)
        ),
        building_density: segmentation?.distribution?.building || 0,
        vegetation_density: segmentation?.distribution?.vegetation || 0,
        road_density: segmentation?.distribution?.road || 0,
        total_objects: objects?.totalCount || 0,
      },
      metadata: {
        city: annotation.city,
        country: annotation.country,
        timestamp: annotation.timestamp,
      },
    };

    samples.push(sample);
  }

  return { samples, count: samples.length };
}

/**
 * Spara dataset i JSONL-format
 */
async function saveDatasets(datasets, outputDir) {
  await fs.mkdir(outputDir, { recursive: true });

  for (const [name, dataset] of Object.entries(datasets)) {
    if (dataset.count === 0) continue;

    const filePath = path.join(outputDir, `${name}.jsonl`);
    const lines = dataset.samples.map(s => JSON.stringify(s)).join('\n');

    await fs.writeFile(filePath, lines + '\n');
    console.log(`[DATASET] Saved ${name}: ${dataset.count} samples`);
  }
}

/**
 * Generera statistik
 */
function generateStats(datasets) {
  const stats = {
    version: DATASET_CONFIG.version,
    createdAt: new Date().toISOString(),
    totalSamples: 0,
    categories: {},
    cities: new Set(),
    countries: new Set(),
  };

  for (const [name, dataset] of Object.entries(datasets)) {
    stats.categories[name] = {
      count: dataset.count,
    };
    stats.totalSamples += dataset.count;

    // Samla unika städer och länder
    for (const sample of dataset.samples) {
      if (sample.metadata?.city) stats.cities.add(sample.metadata.city);
      if (sample.metadata?.country) stats.countries.add(sample.metadata.country);
    }
  }

  stats.cities = Array.from(stats.cities);
  stats.countries = Array.from(stats.countries);

  return stats;
}

/**
 * Splitsa dataset i train/validation/test
 */
async function splitDataset(dataset, splits = DATASET_CONFIG.splits) {
  const shuffled = shuffleArray([...dataset.samples]);

  const trainEnd = Math.floor(shuffled.length * splits.train);
  const valEnd = trainEnd + Math.floor(shuffled.length * splits.validation);

  return {
    train: shuffled.slice(0, trainEnd),
    validation: shuffled.slice(trainEnd, valEnd),
    test: shuffled.slice(valEnd),
  };
}

/**
 * Blanda array
 */
function shuffleArray(array) {
  for (let i = array.length - 1; i > 0; i--) {
    const j = Math.floor(Math.random() * (i + 1));
    [array[i], array[j]] = [array[j], array[i]];
  }
  return array;
}

module.exports = {
  buildDataset,
  buildRoadDataset,
  buildSidewalkDataset,
  buildCrosswalkDataset,
  buildBuildingDataset,
  buildTreeDataset,
  buildUtilityDataset,
  buildTrafficDataset,
  buildSignageDataset,
  buildLightingDataset,
  buildAccessibilityDataset,
  buildCommercialDataset,
  buildDensityDataset,
  splitDataset,
};