tfjs-model-facemesh
Version:
forked from @tensorflow-models/facemesh, used for local deployed tfjs models.
207 lines (206 loc) • 11.4 kB
JavaScript
"use strict";
/**
* @license
* Copyright 2020 Google LLC. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* https://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* =============================================================================
*/
var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
return new (P || (P = Promise))(function (resolve, reject) {
function fulfilled(value) { try { step(generator.next(value)); } catch (e) { reject(e); } }
function rejected(value) { try { step(generator["throw"](value)); } catch (e) { reject(e); } }
function step(result) { result.done ? resolve(result.value) : adopt(result.value).then(fulfilled, rejected); }
step((generator = generator.apply(thisArg, _arguments || [])).next());
});
};
var __generator = (this && this.__generator) || function (thisArg, body) {
var _ = { label: 0, sent: function() { if (t[0] & 1) throw t[1]; return t[1]; }, trys: [], ops: [] }, f, y, t, g;
return g = { next: verb(0), "throw": verb(1), "return": verb(2) }, typeof Symbol === "function" && (g[Symbol.iterator] = function() { return this; }), g;
function verb(n) { return function (v) { return step([n, v]); }; }
function step(op) {
if (f) throw new TypeError("Generator is already executing.");
while (_) try {
if (f = 1, y && (t = op[0] & 2 ? y["return"] : op[0] ? y["throw"] || ((t = y["return"]) && t.call(y), 0) : y.next) && !(t = t.call(y, op[1])).done) return t;
if (y = 0, t) op = [op[0] & 2, t.value];
switch (op[0]) {
case 0: case 1: t = op; break;
case 4: _.label++; return { value: op[1], done: false };
case 5: _.label++; y = op[1]; op = [0]; continue;
case 7: op = _.ops.pop(); _.trys.pop(); continue;
default:
if (!(t = _.trys, t = t.length > 0 && t[t.length - 1]) && (op[0] === 6 || op[0] === 2)) { _ = 0; continue; }
if (op[0] === 3 && (!t || (op[1] > t[0] && op[1] < t[3]))) { _.label = op[1]; break; }
if (op[0] === 6 && _.label < t[1]) { _.label = t[1]; t = op; break; }
if (t && _.label < t[2]) { _.label = t[2]; _.ops.push(op); break; }
if (t[2]) _.ops.pop();
_.trys.pop(); continue;
}
op = body.call(thisArg, _);
} catch (e) { op = [6, e]; y = 0; } finally { f = t = 0; }
if (op[0] & 5) throw op[1]; return { value: op[0] ? op[1] : void 0, done: true };
}
};
var __spreadArrays = (this && this.__spreadArrays) || function () {
for (var s = 0, i = 0, il = arguments.length; i < il; i++) s += arguments[i].length;
for (var r = Array(s), k = 0, i = 0; i < il; i++)
for (var a = arguments[i], j = 0, jl = a.length; j < jl; j++, k++)
r[k] = a[j];
return r;
};
Object.defineProperty(exports, "__esModule", { value: true });
exports.Pipeline = void 0;
var tf = require("@tensorflow/tfjs-core");
var box_1 = require("./box");
var LANDMARKS_COUNT = 468;
var UPDATE_REGION_OF_INTEREST_IOU_THRESHOLD = 0.25;
// The Pipeline coordinates between the bounding box and skeleton models.
var Pipeline = /** @class */ (function () {
function Pipeline(boundingBoxDetector, meshDetector, meshWidth, meshHeight, maxContinuousChecks, maxFaces) {
// An array of facial bounding boxes.
this.regionsOfInterest = [];
this.runsWithoutFaceDetector = 0;
this.boundingBoxDetector = boundingBoxDetector;
this.meshDetector = meshDetector;
this.meshWidth = meshWidth;
this.meshHeight = meshHeight;
this.maxContinuousChecks = maxContinuousChecks;
this.maxFaces = maxFaces;
}
/**
* Returns an array of predictions for each face in the input.
*
* @param input - tensor of shape [1, H, W, 3].
*/
Pipeline.prototype.predict = function (input) {
return __awaiter(this, void 0, void 0, function () {
var returnTensors, annotateFace, _a, boxes, scaleFactor_1, scaledBoxes;
var _this = this;
return __generator(this, function (_b) {
switch (_b.label) {
case 0:
if (!this.shouldUpdateRegionsOfInterest()) return [3 /*break*/, 2];
returnTensors = true;
annotateFace = false;
return [4 /*yield*/, this.boundingBoxDetector.getBoundingBoxes(input, returnTensors, annotateFace)];
case 1:
_a = _b.sent(), boxes = _a.boxes, scaleFactor_1 = _a.scaleFactor;
if (boxes.length === 0) {
scaleFactor_1.dispose();
this.clearAllRegionsOfInterest();
return [2 /*return*/, null];
}
scaledBoxes = boxes.map(function (prediction) { return box_1.enlargeBox(box_1.scaleBoxCoordinates(prediction, scaleFactor_1)); });
boxes.forEach(box_1.disposeBox);
this.updateRegionsOfInterest(scaledBoxes);
this.runsWithoutFaceDetector = 0;
return [3 /*break*/, 3];
case 2:
this.runsWithoutFaceDetector++;
_b.label = 3;
case 3: return [2 /*return*/, tf.tidy(function () {
return _this.regionsOfInterest.map(function (box, i) {
var face = box_1.cutBoxFromImageAndResize(box, input, [
_this.meshHeight, _this.meshWidth
]).div(255);
// The first returned tensor represents facial contours, which are
// included in the coordinates.
var _a = _this.meshDetector.predict(face), flag = _a[1], coords = _a[2];
var coordsReshaped = tf.reshape(coords, [-1, 3]);
var normalizedBox = tf.div(box_1.getBoxSize(box), [_this.meshWidth, _this.meshHeight]);
var scaledCoords = tf.mul(coordsReshaped, normalizedBox.concat(tf.tensor2d([1], [1, 1]), 1))
.add(box.startPoint.concat(tf.tensor2d([0], [1, 1]), 1));
var landmarksBox = _this.calculateLandmarksBoundingBox(scaledCoords);
var previousBox = _this.regionsOfInterest[i];
box_1.disposeBox(previousBox);
_this.regionsOfInterest[i] = landmarksBox;
var prediction = {
coords: coordsReshaped,
scaledCoords: scaledCoords,
box: landmarksBox,
flag: flag.squeeze()
};
return prediction;
});
})];
}
});
});
};
// Updates regions of interest if the intersection over union between
// the incoming and previous regions falls below a threshold.
Pipeline.prototype.updateRegionsOfInterest = function (boxes) {
for (var i = 0; i < boxes.length; i++) {
var box = boxes[i];
var previousBox = this.regionsOfInterest[i];
var iou = 0;
if (previousBox && previousBox.startPoint) {
// Computing IOU on the CPU for performance.
// Using arraySync() rather than await array() because the tensors are
// very small, so it's not worth the overhead to call await array().
var _a = box.startEndTensor.arraySync()[0], boxStartX = _a[0], boxStartY = _a[1], boxEndX = _a[2], boxEndY = _a[3];
var _b = previousBox.startEndTensor.arraySync()[0], previousBoxStartX = _b[0], previousBoxStartY = _b[1], previousBoxEndX = _b[2], previousBoxEndY = _b[3];
var xStartMax = Math.max(boxStartX, previousBoxStartX);
var yStartMax = Math.max(boxStartY, previousBoxStartY);
var xEndMin = Math.min(boxEndX, previousBoxEndX);
var yEndMin = Math.min(boxEndY, previousBoxEndY);
var intersection = (xEndMin - xStartMax) * (yEndMin - yStartMax);
var boxArea = (boxEndX - boxStartX) * (boxEndY - boxStartY);
var previousBoxArea = (previousBoxEndX - previousBoxStartX) *
(previousBoxEndY - boxStartY);
iou = intersection / (boxArea + previousBoxArea - intersection);
}
if (iou > UPDATE_REGION_OF_INTEREST_IOU_THRESHOLD) {
box_1.disposeBox(box);
}
else {
this.regionsOfInterest[i] = box;
box_1.disposeBox(previousBox);
}
}
for (var i = boxes.length; i < this.regionsOfInterest.length; i++) {
box_1.disposeBox(this.regionsOfInterest[i]);
}
this.regionsOfInterest = this.regionsOfInterest.slice(0, boxes.length);
};
Pipeline.prototype.clearRegionOfInterest = function (index) {
if (this.regionsOfInterest[index] != null) {
box_1.disposeBox(this.regionsOfInterest[index]);
this.regionsOfInterest = __spreadArrays(this.regionsOfInterest.slice(0, index), this.regionsOfInterest.slice(index + 1));
}
};
Pipeline.prototype.clearAllRegionsOfInterest = function () {
for (var i = 0; i < this.regionsOfInterest.length; i++) {
box_1.disposeBox(this.regionsOfInterest[i]);
}
this.regionsOfInterest = [];
};
Pipeline.prototype.shouldUpdateRegionsOfInterest = function () {
var roisCount = this.regionsOfInterest.length;
var noROIs = roisCount === 0;
if (this.maxFaces === 1 || noROIs) {
return noROIs;
}
return roisCount !== this.maxFaces &&
this.runsWithoutFaceDetector >= this.maxContinuousChecks;
};
Pipeline.prototype.calculateLandmarksBoundingBox = function (landmarks) {
var xs = landmarks.slice([0, 0], [LANDMARKS_COUNT, 1]);
var ys = landmarks.slice([0, 1], [LANDMARKS_COUNT, 1]);
var boxMinMax = tf.stack([xs.min(), ys.min(), xs.max(), ys.max()]);
var box = box_1.createBox(boxMinMax.expandDims(0));
return box_1.enlargeBox(box);
};
return Pipeline;
}());
exports.Pipeline = Pipeline;