Files
scene-actor-extraction/src/gallery/gallery_builder.cpp
T
dtourolle d98dc2855a refactor(bench): SuperHero replaces Road to Bali as the reference film
Bali was chosen because the TRECVID DVU set ships character mugshots, but
its reference crops are unusable at scale: median detected face 27 px
against a 69 px maximum, so every reference was upscaled 4x or more past
what the embedder was trained for (AR-011). A 66 px floor left 2 of 69
references; no threshold exists that both keeps the faces in distribution
and leaves enough of them to calibrate.

SuperHero is 69 px median and 241 px max. Its gallery builds at a 66 px
floor with 14 references over 5 characters, and calibrates on its own
(a=15.2867 b=-4.98633, 100% train accuracy) instead of borrowing constants.

Measured on the fused 17-minute film, one stream rather than per-scene
clips so presence windows cross real scene boundaries as SR-002 intends:
precision 1.00, recall 0.65, F1 0.79 — 13 true positives, 0 false
positives, 7 misses. Every out-of-gallery character was declined rather
than forced onto a nearest match. The misses are the short scenes (14 s,
38 s, 27 s), consistent with per-track accumulation needing sightings.

- build_gallery gains --min-face-px, filtering the *detected face* rather
  than the crop. The DVU images are scene crops, not mugshots, so crop
  dimensions say nothing about face scale. A poisoned reference is
  permanent in a way a bad frame is not: it corrupts every future match
  against that identity.
- scripts/fetch_dvu.sh fetches mugshots, scene graphs and segmentation for
  any DVU film. NIST names the same film three different ways, so KG_DIR
  and KG_FILE are overridable rather than derived. This exists as a script
  because the first copy of this data was assembled ad hoc in /tmp and was
  lost with it, taking the working gallery along.
- Replay fixtures move to the artifact registry: push/pull_artifacts.sh
  gain a replay-fixtures target, and tests/fixtures/dumps/.gitignore keeps
  them out of git. superhero.h5 is ~9 MB and regenerating it needs the
  film, the models and a GPU — none of which CI has. The gallery ships
  with the dumps, since a dump only replays against the gallery it was
  produced with.
- AR-012 and AR-013 coverage is ported onto the new fixture rather than
  dropped with the Bali cases: 12369 assertions, up from 7991, since the
  film is an order of magnitude larger than the clips.

Suite: 15679 assertions, 101 test cases.

TRACES: AR-011, AR-012, AR-013 | VR-001, VR-005 | SR-002
2026-08-04 13:49:11 +02:00

146 lines
5.8 KiB
C++

#include "gallery_builder.hpp"
#include "config.hpp"
#include "embedder_stamp.hpp"
#include "face_utils.hpp"
#include "inference/face_detector.hpp"
#include "inference/face_embedder.hpp"
#include <opencv2/imgcodecs.hpp>
#include <algorithm>
#include <filesystem>
#include <iostream>
#include <stdexcept>
#include <string>
namespace fs = std::filesystem;
// ── Parse "nm0000093_Brad_Pitt" → ("nm0000093", "Brad Pitt") ─────────────────
static std::pair<std::string, std::string> parse_dir_name(const std::string& dirname) {
auto pos = dirname.find('_');
if (pos == std::string::npos) return {dirname, dirname};
std::string imdb_id = dirname.substr(0, pos);
std::string raw = dirname.substr(pos + 1);
std::string name;
name.reserve(raw.size());
for (char c : raw)
name += (c == '_' ? ' ' : c);
return {imdb_id, name};
}
// ── Public API ────────────────────────────────────────────────────────────────
ActorGallery build_gallery(const BuildConfig& cfg) {
Config icfg;
icfg.detector_model = cfg.detector_model;
icfg.arcface_model = cfg.arcface_model;
icfg.detector_conf = cfg.detector_conf;
icfg.detector_nms = cfg.detector_nms;
auto decoder = make_face_detector(icfg);
auto arcface = make_face_embedder(icfg);
ActorGallery gallery;
/// TRACES: GR-004 | SR-001
// Stamp before the first embedding exists, so there is no window in which a
// gallery holds vectors without recording what produced them.
gallery.embedder = make_embedder_stamp(cfg.arcface_model);
std::cerr << "[build_gallery] embedder: " << gallery.embedder.describe() << "\n";
for (const auto& actor_dir : fs::directory_iterator(cfg.gallery_root)) {
if (!actor_dir.is_directory()) continue;
auto [imdb_id, name] = parse_dir_name(actor_dir.path().filename().string());
std::cerr << "[build_gallery] " << name << " (" << imdb_id << ")\n";
ActorGallery::Actor actor;
actor.imdb_id = imdb_id;
actor.name = name;
static const std::vector<std::string> kExts{".jpg", ".jpeg", ".png", ".webp"};
for (const auto& img_file : fs::directory_iterator(actor_dir.path())) {
if (!img_file.is_regular_file()) continue;
std::string ext = img_file.path().extension().string();
std::transform(ext.begin(), ext.end(), ext.begin(), ::tolower);
if (std::find(kExts.begin(), kExts.end(), ext) == kExts.end()) continue;
cv::Mat img = cv::imread(img_file.path().string());
if (img.empty()) {
std::cerr << " [skip] cannot read " << img_file.path().filename() << "\n";
continue;
}
if (cfg.max_side > 0) {
const int big = std::max(img.cols, img.rows);
if (big > cfg.max_side) {
const double s = static_cast<double>(cfg.max_side) / big;
cv::resize(img, img, {}, s, s, cv::INTER_AREA);
}
}
auto faces = decoder->detect(img);
if (faces.empty()) {
std::cerr << " [skip] no face: " << img_file.path().filename() << "\n";
continue;
}
if (faces.size() > 1) {
std::cerr << " [warn] " << faces.size() << " faces, using highest confidence: "
<< img_file.path().filename() << "\n";
}
const auto& best = *std::max_element(
faces.begin(), faces.end(),
[](const DetectedFace& a, const DetectedFace& b) {
return a.confidence < b.confidence;
});
// Reject faces too small to embed honestly.
//
// The reference images are crops cut from the film, not mugshots, so
// the detected face can be a small fraction of the image. Upscaling a
// 30 px face to ArcFace's 112x112 feeds the model an input it was
// never trained for, and it answers with a confident, plausible,
// wrong embedding.
//
// At inference that costs one frame. Here it is permanent: a poisoned
// reference sits in the gallery and corrupts every future match
// against that character, which is exactly the kind of error that is
// invisible without a study that should not have been needed.
if (cfg.min_face_px > 0.f) {
const float side = std::min(best.bbox.width, best.bbox.height);
if (side < cfg.min_face_px) {
std::cerr << " [skip] face " << side << "px < " << cfg.min_face_px
<< "px: " << img_file.path().filename() << "\n";
continue;
}
}
cv::Mat crop = align_face(img, best.landmarks);
if (crop.empty()) {
std::cerr << " [skip] alignment failed: " << img_file.path().filename() << "\n";
continue;
}
Embedding emb = arcface->embed_one(crop);
actor.embeddings.push_back(emb);
actor.source_images.push_back(img_file.path().filename().string());
std::cerr << " [ok] " << img_file.path().filename()
<< " conf=" << best.confidence << "\n";
}
if (actor.embeddings.empty()) {
std::cerr << " [warn] no valid embeddings for " << name << " — skipped\n";
continue;
}
std::cerr << " → " << actor.embeddings.size() << " embeddings\n";
gallery.actors.push_back(std::move(actor));
}
std::cerr << "[build_gallery] total: " << gallery.actors.size() << " actors\n";
return gallery;
}