DarkRoom/ui/dr-ui/examples/face_index.rs

//! The face-indexing batch job, off the GUI.
//!
//! Checks every library image for a face-detection run marker, and optionally
//! indexes whatever is missing one.
//!
//!   cargo run -p dr-ui --example face_index -- CATALOG.db THUMBS_DIR [--run DET.onnx EMB.onnx]
//!   cargo run -p dr-ui --example face_index -- CATALOG.db THUMBS_DIR --cluster
//!
//! # Why this exists beside the button in the Identity screen
//!
//! Indexing a real library is hours of work (docs/faces.md §12.2), and the
//! cases where that is worth starting — an overnight pass, a fresh import, a
//! machine left running — are exactly the ones where holding a window open is
//! the wrong shape. The check half is useful on its own: it is cheap, it
//! answers "has face recognition been over all of this", and it distinguishes
//! *not yet indexed* from *waiting on a proxy*, which are different problems
//! with different fixes.
//!
//! The models must have had their input dims frozen first; see
//! `tools/fix-face-model-shapes.sh`.

use std::path::PathBuf;

use dr_catalog::Catalog;
use dr_thumbs::ThumbStore;
use dr_ui::faces::{self, FaceSweepMessage};

const MODEL_ID: &str = "w600k_mbf";

fn main() {
    env_logger::init();

    let args: Vec<String> = std::env::args().skip(1).collect();
    if args.len() < 2 {
        eprintln!(
            "usage: face_index CATALOG.db THUMBS_DIR [--run DETECTOR.onnx EMBEDDER.onnx]\n\
             \n\
             With no --run this only reports; nothing is written.\n\
             --cluster groups what is indexed; --tune compares thresholds without writing.\n\
             --quality DET.onnx EMB.onnx reports face size and sharpness, also without writing."
        );
        std::process::exit(2);
    }

    let catalog_path = PathBuf::from(&args[0]);
    let store_dir = PathBuf::from(&args[1]);

    let catalog = match Catalog::open(&catalog_path) {
        Ok(c) => c,
        Err(e) => {
            eprintln!("cannot open catalog {}: {e}", catalog_path.display());
            std::process::exit(1);
        }
    };
    let store = match ThumbStore::open(&store_dir) {
        Ok(s) => s,
        Err(e) => {
            eprintln!("cannot open thumbnail store {}: {e}", store_dir.display());
            std::process::exit(1);
        }
    };

    // The registry a device without the eye models would run: what this
    // example counts as work is what the app would.
    let repairs = dr_ui::repairs::registry(
        dr_ui::repairs::Scope::Outstanding,
        MODEL_ID,
        dr_types::FaceDetector::Scrfd500m,
        dr_ui::repairs::Capabilities {
            gpu: true,
            face_models: true,
            eye_models: false,
        },
    );
    let audit = match faces::audit(&catalog, &store, MODEL_ID, &repairs) {
        Ok(a) => a,
        Err(e) => {
            eprintln!("coverage check failed: {e}");
            std::process::exit(1);
        }
    };

    println!("model      {MODEL_ID}");
    println!("images     {}", audit.coverage.images);
    println!(
        "indexed    {} ({:.1}%)",
        audit.coverage.indexed,
        audit.coverage.fraction() * 100.0
    );
    println!("faces      {}", audit.coverage.faces);
    println!(
        "no faces   {}  (indexed, nothing found — the common case)",
        audit.coverage.without_faces
    );
    println!("outstanding {}", audit.coverage.outstanding());
    println!("  ready to index  {}", audit.ready);
    println!("  awaiting proxy  {}", audit.awaiting_proxy);

    // Grouping is a separate step from indexing on purpose: it is a
    // whole-library operation over the embeddings detection produced, and it is
    // worth running *after* a sweep rather than during one (catalog.md §10.2).
    if args.iter().any(|a| a == "--cluster") {
        // The engine's own defaults, not this device's settings file: a batch
        // job run over a library on a server has no business inheriting the
        // dials somebody moved on their laptop.
        let grouping = dr_types::settings::FaceSettings::default();
        match dr_ui::faces::recluster(&catalog, MODEL_ID, &grouping) {
            Ok((suggested, created)) => {
                println!("\nclustering: {suggested} suggestion(s), {created} new group(s)");
                report_people(&catalog);
            }
            Err(e) => {
                eprintln!("clustering failed: {e}");
                std::process::exit(1);
            }
        }
        return;
    }

    // Tuning, and deliberately read-only: it answers "what would this
    // threshold do to my library" without writing a single suggestion, which
    // is the only way to compare several without each one polluting the next.
    if args.iter().any(|a| a == "--tune") {
        tune_thresholds(&catalog);
        return;
    }

    // Quality tuning, and read-only like `--tune`. Runs the real detector over
    // real proxies with **both gates disabled**, so the distribution it prints
    // is of everything the detector finds rather than of what survives the
    // current settings — which is the only way to see what a threshold would
    // actually remove.
    if let Some(i) = args.iter().position(|a| a == "--quality") {
        let (Some(detector), Some(embedder)) = (args.get(i + 1), args.get(i + 2)) else {
            eprintln!("--quality needs both a detector and an embedder");
            std::process::exit(2);
        };
        report_quality(
            &catalog,
            &store,
            std::path::Path::new(detector),
            std::path::Path::new(embedder),
        );
        return;
    }

    let run = args.iter().position(|a| a == "--run");
    let Some(i) = run else {
        if audit.coverage.is_complete() {
            println!("\nnothing outstanding.");
        } else {
            println!("\npass --run DETECTOR.onnx EMBEDDER.onnx to index the outstanding images.");
        }
        return;
    };

    let (Some(detector), Some(embedder)) = (args.get(i + 1), args.get(i + 2)) else {
        eprintln!("--run needs both a detector and an embedder");
        std::process::exit(2);
    };

    if audit.ready == 0 {
        println!("\nnothing ready to index.");
        if audit.awaiting_proxy > 0 {
            // Worth saying plainly: running this again will not help, because
            // the blocker is in the thumbnail store rather than here.
            println!(
                "{} image(s) are waiting on a proxy — run the thumbnail sweep first.",
                audit.awaiting_proxy
            );
        }
        return;
    }

    // Say it here rather than letting the pass return an empty result and
    // print "done: 0 image(s)". That is the shape of report this whole change
    // exists to stop producing.
    if faces::FACE_TIER.edge() < dr_face::MIN_CROP_EDGE {
        println!(
            "\n--run cannot index from the local store any more.\n\
             \n\
             Face crops must be sampled from a buffer longer than {}px and the\n\
             store's largest tier is {}px, so every image here would be refused.\n\
             The measurement behind that floor is docs/faces.md §7b.\n\
             \n\
             Index from the app's Identity screen instead: that pass renders at\n\
             native resolution, which is what the crop needs.",
            dr_face::MIN_CROP_EDGE - 1,
            faces::FACE_TIER.edge(),
        );
        return;
    }

    println!("\nindexing {} image(s)…", audit.ready);
    let rx = faces::spawn_store_face_sweep(
        catalog_path,
        store_dir,
        dr_ui::FaceModelPaths {
            detector: PathBuf::from(detector),
            embedder: PathBuf::from(embedder),
            // A measurement tool for the detector and embedder; the eye
            // models are the library pass's business.
            eyes: None,
        },
        MODEL_ID.to_string(),
        dr_face::DetectOptions::default(),
    );

    let mut seen = 0usize;
    let mut total = 0usize;
    let start = std::time::Instant::now();
    for msg in rx {
        match msg {
            FaceSweepMessage::Total(n) => total = n,
            FaceSweepMessage::Indexed { faces, .. } => {
                seen += 1;
                // One line per image would be thousands of lines; one per
                // twenty-five is enough to see it moving and to estimate.
                if seen.is_multiple_of(25) || faces > 0 {
                    let rate = seen as f64 / start.elapsed().as_secs_f64().max(1e-6);
                    println!(
                        "  {seen}/{total}  {:.2} img/s  ~{:.0} min left",
                        rate,
                        (total.saturating_sub(seen)) as f64 / rate.max(1e-6) / 60.0
                    );
                }
            }
            FaceSweepMessage::Failed { images } => seen += images,
            FaceSweepMessage::Finished {
                images,
                faces,
                failed,
            } => {
                println!(
                    "\ndone: {images} image(s), {faces} face(s), {failed} failed, in {:.0}s",
                    start.elapsed().as_secs_f64()
                );
            }
        }
    }

    if let Ok(a) = faces::audit(&catalog, &store, MODEL_ID, &repairs) {
        println!("{}", a.summary());
    }
    println!("\nrun again with --cluster to group these faces into people.");
}

/// What the clustering proposed, largest group first.
fn report_people(catalog: &Catalog) {
    let Ok(people) = dr_catalog::faces::people(catalog.connection()) else {
        return;
    };
    if people.is_empty() {
        println!("no groups — too few faces, or none similar enough to group.");
        return;
    }
    println!("\n{} group(s):", people.len());
    for p in people.iter().take(30) {
        let name = if p.name.is_empty() {
            "(unnamed)".to_string()
        } else {
            p.name.clone()
        };
        println!(
            "  {name:<24} {} confirmed, {} suggested",
            p.confirmed_faces, p.suggested_faces
        );
    }
    if people.len() > 30 {
        println!("  … and {} more", people.len() - 30);
    }
}

/// What several merge thresholds would each do to this library.
///
/// The default 0.9 is a *probability*, and the cosine it lands on depends on
/// the calibration — so "is 0.9 too tight" is not a question anyone can answer
/// from the number alone. This runs the real clusterer over the real
/// embeddings at a range of thresholds and prints what each one produces, which
/// is the only honest way to choose.
///
/// Nothing is written. Run it, read the table, then pass the number you want.
fn tune_thresholds(catalog: &Catalog) {
    use dr_catalog::faces;

    let conn = catalog.connection();
    let cal = match faces::calibration(conn, MODEL_ID) {
        Ok(Some((c, _))) => c,
        _ => dr_face::Calibration::default(),
    };

    let stored = match faces::embeddings(conn, MODEL_ID) {
        Ok(s) => s,
        Err(e) => {
            eprintln!("cannot read embeddings: {e}");
            std::process::exit(1);
        }
    };
    if stored.is_empty() {
        println!("no faces indexed yet — nothing to tune.");
        return;
    }

    let model = dr_face::ModelId::new(MODEL_ID.to_string());
    let mut candidates = Vec::with_capacity(stored.len());
    for f in stored {
        let Some(emb) = dr_face::Embedding::from_f16_bytes(model.clone(), &f.embedding) else {
            continue;
        };
        candidates.push(dr_face::Candidate {
            face: f.face.0,
            image: f.image.0,
            embedding: emb.v.to_vec(),
            crop_px: f.crop_px,
            quality: f.quality,
            confirmed_person: None,
        });
    }

    println!(
        "\n{} face(s), calibration valid: {}",
        candidates.len(),
        cal.valid
    );
    println!(
        "\n{:>6}  {:>7}  {:>7}  {:>7}  {:>7}  {:>8}  {:>7}",
        "P", "cosine", "groups", "grouped", "largest", "in groups", "time"
    );
    println!("{}", "-".repeat(60));

    for p in [0.99_f32, 0.97, 0.95, 0.9, 0.85, 0.8, 0.75, 0.7, 0.6, 0.5] {
        let start = std::time::Instant::now();
        let clusters = dr_face::cluster(&candidates, &cal, p);
        let elapsed = start.elapsed();

        // A group of one is not a person, and `recluster` discards those, so
        // the interesting figures count only the real groups.
        let real: Vec<_> = clusters.iter().filter(|c| c.members.len() >= 2).collect();
        let grouped: usize = real.iter().map(|c| c.members.len()).sum();
        let largest = real.first().map(|c| c.members.len()).unwrap_or(0);

        println!(
            "{p:>6.2}  {:>7.3}  {:>7}  {:>7}  {:>7}  {:>7.0}%  {:>6.2}s",
            cal.boundary_at(p, 150.0, 0.0),
            real.len(),
            grouped,
            largest,
            100.0 * grouped as f64 / candidates.len() as f64,
            elapsed.as_secs_f64(),
        );
    }

    println!(
        "\nA looser threshold makes bigger groups and merges people who are not\n\
         the same; a tighter one splits one person across several. The largest\n\
         group is the tell: when it starts growing much faster than the rest,\n\
         identities are being welded together."
    );
}

/// How many images to sample for the quality report.
///
/// Enough for the distribution to settle, few enough to finish while the user
/// is watching: detection is ~100ms an image, so this is a couple of minutes.
const QUALITY_SAMPLE: usize = 600;

/// What the detector finds, before either quality gate is applied.
///
/// The two floors — face size and sharpness — are not independent: a face
/// smaller than the embedder's 112-pixel input was upsampled to reach it, and
/// upsampling invents no edges, so small faces score low on sharpness even when
/// the original was crisp. Choosing either number without seeing the other is
/// how you end up with one gate doing nothing and the other doing too much.
///
/// So this prints them together, over the real library, with nothing filtered.
fn report_quality(
    catalog: &Catalog,
    store: &ThumbStore,
    detector: &std::path::Path,
    embedder: &std::path::Path,
) {
    let mut det = match dr_face::Detector::from_path(detector) {
        Ok(d) => d,
        Err(e) => {
            eprintln!("cannot load the detector: {e}");
            std::process::exit(1);
        }
    };
    // Loaded but unused: the point is to fail here, before a two-minute scan,
    // if the pair the user passed is not the pair indexing would use.
    if let Err(e) =
        dr_face::Embedder::from_path(embedder, dr_face::ModelId::new(MODEL_ID.to_string()))
    {
        eprintln!("cannot load the embedder: {e}");
        std::process::exit(1);
    }

    // Everything the detector can find: no size floor, no sharpness floor.
    let options = dr_face::DetectOptions {
        min_face_px: 0.0,
        min_source_px: 0.0,
        min_sharpness: 0.0,
        ..Default::default()
    };

    let mut stmt = match catalog.connection().prepare(
        "SELECT r.file_id FROM remote r
           JOIN images i ON i.id = r.image_id
          WHERE r.file_id IS NOT NULL AND i.trashed_at IS NULL
          ORDER BY i.id",
    ) {
        Ok(s) => s,
        Err(e) => {
            eprintln!("cannot list images: {e}");
            std::process::exit(1);
        }
    };
    let file_ids: Vec<u64> = stmt
        .query_map([], |r| r.get::<_, i64>(0))
        .into_iter()
        .flatten()
        .filter_map(Result::ok)
        .map(|v| v as u64)
        .filter(|id| store.contains(*id, faces::FACE_TIER))
        .take(QUALITY_SAMPLE)
        .collect();

    if file_ids.is_empty() {
        println!("no proxies on disk to measure — browse the library first.");
        return;
    }
    println!("\nmeasuring {} image(s)…", file_ids.len());

    // (source_px, sharpness) per detected face.
    let mut found: Vec<(f32, f32)> = Vec::new();
    let mut images = 0usize;
    for id in &file_ids {
        let Ok(Some(thumb)) = store.get(*id, faces::FACE_TIER) else {
            continue;
        };
        let Ok((w, h, rgba)) = dr_thumbs::codec::decode_rgba(&thumb.bytes) else {
            continue;
        };
        let rgb: Vec<f32> = rgba
            .chunks_exact(4)
            .flat_map(|p| {
                [
                    p[0] as f32 / 255.0,
                    p[1] as f32 / 255.0,
                    p[2] as f32 / 255.0,
                ]
            })
            .collect();

        let Ok(dets) = det.detect(&rgb, w as usize, h as usize, &options) else {
            continue;
        };
        images += 1;
        for d in &dets {
            if let Some(a) = dr_face::warp(&rgb, w as usize, h as usize, &d.landmarks) {
                found.push((a.source_px(), a.sharpness()));
            }
        }
        if images.is_multiple_of(50) {
            println!(
                "  {images}/{} images, {} face(s)",
                file_ids.len(),
                found.len()
            );
        }
    }

    if found.is_empty() {
        println!("no faces found in the sample.");
        return;
    }

    let pct = |v: &mut Vec<f32>, p: f64| -> f32 {
        v.sort_by(|a, b| a.total_cmp(b));
        v[(((v.len() - 1) as f64) * p) as usize]
    };
    let mut sizes: Vec<f32> = found.iter().map(|f| f.0).collect();
    let mut sharps: Vec<f32> = found.iter().map(|f| f.1).collect();

    println!("\n{} face(s) in {images} image(s)\n", found.len());
    println!(
        "{:>12}  {:>8}  {:>10}",
        "percentile", "size px", "sharpness"
    );
    println!("{}", "-".repeat(34));
    for p in [0.01, 0.05, 0.10, 0.25, 0.50, 0.75, 0.90, 0.99] {
        println!(
            "{:>11.0}%  {:>8.0}  {:>10.4}",
            p * 100.0,
            pct(&mut sizes, p),
            pct(&mut sharps, p)
        );
    }

    // What each candidate pair would remove. Cumulative, because the gates are
    // applied together and their overlap is the whole question.
    println!(
        "\n{:>8}  {:>10}  {:>9}  {:>9}  {:>9}",
        "min crop", "min sharp", "size cut", "blur cut", "kept"
    );
    println!("{}", "-".repeat(52));
    for (min_px, min_sharp) in [
        (0.0_f32, 0.0_f32),
        (32.0, 0.0),
        (32.0, 0.002),
        (32.0, 0.005),
        (32.0, 0.010),
        (32.0, 0.020),
        (48.0, 0.005),
        (64.0, 0.010),
        (64.0, 0.020),
    ] {
        let by_size = found.iter().filter(|f| f.0 < min_px).count();
        let by_blur = found
            .iter()
            .filter(|f| f.0 >= min_px && f.1 < min_sharp)
            .count();
        let kept = found.len() - by_size - by_blur;
        println!(
            "{min_px:>8.0}  {min_sharp:>10.3}  {:>8.0}%  {:>8.0}%  {:>8.0}%",
            100.0 * by_size as f64 / found.len() as f64,
            100.0 * by_blur as f64 / found.len() as f64,
            100.0 * kept as f64 / found.len() as f64,
        );
    }

    println!(
        "\n`size cut` is what the size floor removes; `blur cut` is what the\n\
         sharpness floor removes *of what the size floor left*, so the two\n\
         columns do not double-count. A sharpness floor that cuts almost\n\
         nothing once the size floor is in place is a floor that is not\n\
         earning its place."
    );
}