face-api/build/ssdMobilenetv1/mobileNetV1.js

51 lines
1.8 KiB
JavaScript

import * as tf from '@tensorflow/tfjs-core';
import { pointwiseConvLayer } from './pointwiseConvLayer';
const epsilon = 0.0010000000474974513;
function depthwiseConvLayer(x, params, strides) {
return tf.tidy(() => {
let out = tf.depthwiseConv2d(x, params.filters, strides, 'same');
out = tf.batchNorm(out, params.batch_norm_mean, params.batch_norm_variance, params.batch_norm_offset, params.batch_norm_scale, epsilon);
return tf.clipByValue(out, 0, 6);
});
}
function getStridesForLayerIdx(layerIdx) {
return [2, 4, 6, 12].some(idx => idx === layerIdx) ? [2, 2] : [1, 1];
}
export function mobileNetV1(x, params) {
return tf.tidy(() => {
let conv11 = null;
let out = pointwiseConvLayer(x, params.conv_0, [2, 2]);
const convPairParams = [
params.conv_1,
params.conv_2,
params.conv_3,
params.conv_4,
params.conv_5,
params.conv_6,
params.conv_7,
params.conv_8,
params.conv_9,
params.conv_10,
params.conv_11,
params.conv_12,
params.conv_13
];
convPairParams.forEach((param, i) => {
const layerIdx = i + 1;
const depthwiseConvStrides = getStridesForLayerIdx(layerIdx);
out = depthwiseConvLayer(out, param.depthwise_conv, depthwiseConvStrides);
out = pointwiseConvLayer(out, param.pointwise_conv, [1, 1]);
if (layerIdx === 11) {
conv11 = out;
}
});
if (conv11 === null) {
throw new Error('mobileNetV1 - output of conv layer 11 is null');
}
return {
out,
conv11: conv11
};
});
}
//# sourceMappingURL=mobileNetV1.js.map