face-api.js
Version:
JavaScript API for face detection and face recognition in the browser with tensorflow.js
70 lines • 3.72 kB
JavaScript
;
Object.defineProperty(exports, "__esModule", { value: true });
var tslib_1 = require("tslib");
var tf = require("@tensorflow/tfjs-core");
var ops_1 = require("../ops");
var extractImagePatches_1 = require("./extractImagePatches");
var MtcnnBox_1 = require("./MtcnnBox");
var RNet_1 = require("./RNet");
function stage2(img, inputBoxes, scoreThreshold, params, stats) {
return tslib_1.__awaiter(this, void 0, void 0, function () {
var ts, rnetInputs, rnetOuts, scoresTensor, scores, _a, _b, indices, filteredBoxes, filteredScores, finalBoxes, finalScores, indicesNms, regions_1;
return tslib_1.__generator(this, function (_c) {
switch (_c.label) {
case 0:
ts = Date.now();
return [4 /*yield*/, extractImagePatches_1.extractImagePatches(img, inputBoxes, { width: 24, height: 24 })];
case 1:
rnetInputs = _c.sent();
stats.stage2_extractImagePatches = Date.now() - ts;
ts = Date.now();
rnetOuts = rnetInputs.map(function (rnetInput) {
var out = RNet_1.RNet(rnetInput, params);
rnetInput.dispose();
return out;
});
stats.stage2_rnet = Date.now() - ts;
scoresTensor = rnetOuts.length > 1
? tf.concat(rnetOuts.map(function (out) { return out.scores; }))
: rnetOuts[0].scores;
_b = (_a = Array).from;
return [4 /*yield*/, scoresTensor.data()];
case 2:
scores = _b.apply(_a, [_c.sent()]);
scoresTensor.dispose();
indices = scores
.map(function (score, idx) { return ({ score: score, idx: idx }); })
.filter(function (c) { return c.score > scoreThreshold; })
.map(function (_a) {
var idx = _a.idx;
return idx;
});
filteredBoxes = indices.map(function (idx) { return inputBoxes[idx]; });
filteredScores = indices.map(function (idx) { return scores[idx]; });
finalBoxes = [];
finalScores = [];
if (filteredBoxes.length > 0) {
ts = Date.now();
indicesNms = ops_1.nonMaxSuppression(filteredBoxes, filteredScores, 0.7);
stats.stage2_nms = Date.now() - ts;
regions_1 = indicesNms.map(function (idx) {
var regionsData = rnetOuts[indices[idx]].regions.arraySync();
return new MtcnnBox_1.MtcnnBox(regionsData[0][0], regionsData[0][1], regionsData[0][2], regionsData[0][3]);
});
finalScores = indicesNms.map(function (idx) { return filteredScores[idx]; });
finalBoxes = indicesNms.map(function (idx, i) { return filteredBoxes[idx].calibrate(regions_1[i]); });
}
rnetOuts.forEach(function (t) {
t.regions.dispose();
t.scores.dispose();
});
return [2 /*return*/, {
boxes: finalBoxes,
scores: finalScores
}];
}
});
});
}
exports.stage2 = stage2;
//# sourceMappingURL=stage2.js.map