forked from tensorflow/tfjs
-
Notifications
You must be signed in to change notification settings - Fork 5
/
ResizeBicubic.ts
131 lines (113 loc) · 5.09 KB
/
ResizeBicubic.ts
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
/**
* @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
*
* http://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.
* =============================================================================
*/
import {KernelConfig, KernelFunc, ResizeBicubic, ResizeBicubicAttrs, ResizeBicubicInputs, TensorInfo, TypedArray, util} from '@tensorflow/tfjs-core';
import {MathBackendCPU} from '../backend_cpu';
import {assertNotComplex} from '../cpu_util';
export function resizeBicubic(args: {
inputs: ResizeBicubicInputs,
backend: MathBackendCPU,
attrs: ResizeBicubicAttrs
}): TensorInfo {
const {inputs, backend, attrs} = args;
const {images} = inputs;
const {alignCorners, halfPixelCenters, size} = attrs;
console.log(`alignCorners ${alignCorners}`)
console.log(`halfPixelCenters ${halfPixelCenters}`)
assertNotComplex(images, 'resizeBicubic');
const imagesStrides = util.computeStrides(images.shape);
const [newHeight, newWidth] = size;
const [batch, oldHeight, oldWidth, numChannels] = images.shape;
const xValues = backend.data.get(images.dataId).values as TypedArray;
console.log(`xValues ${xValues}`)
console.log(`imagesStrides ${imagesStrides}`)
function indexToCoordinates(point: number, imagesStrides: number[]): number[] {
const rowPoint = Math.floor(point / imagesStrides[1])
const colPoint = point % imagesStrides[1]
return [rowPoint, colPoint]
}
function coordinatesToIndex(row: number, col:number, imagesStrides: number[]) {
return (row * imagesStrides[1]) + col
}
function padInputImage(img: TypedArray): Float32Array {
const pad = 2;
const paddedMat = new Float32Array(util.sizeFromShape([batch, oldHeight + pad*2, oldWidth + pad*2, numChannels]));
const paddedStrides = util.computeStrides([batch, oldHeight + pad*2, oldWidth + pad*2, numChannels]);
const [imgRow, imgCol] = [(imagesStrides[0] / imagesStrides[1]), imagesStrides[1]];
const maxLen = paddedMat.length;
for (let i = 0; i < maxLen; i++) {
const [row, col] = indexToCoordinates(i, paddedStrides);
const topRow = 0;
const bottomRow = imgRow - 1;
const leftCol = 0;
const rightCol = imgCol - 1;
const currentCol = col - pad;
const currentRow = row - pad;
if (row < pad) { // case: top
if (col < pad) { // case: top left
paddedMat[i] = img[coordinatesToIndex(topRow, leftCol, imagesStrides)]; // top left from original img
}
else if (col > imgCol + pad - 1) { // case: top right
paddedMat[i] = img[coordinatesToIndex(topRow, rightCol, imagesStrides)]; // top right from original img
}
else { // case: top, not corner
paddedMat[i] = img[coordinatesToIndex(topRow, currentCol, imagesStrides)]; // top from original img
}
}
else if (row > imgRow + pad - 1) { // case: bottom
if (col < pad) { // case: bottom left
paddedMat[i] = img[coordinatesToIndex(bottomRow, leftCol, imagesStrides)]; // bottom left from original img
}
else if (col > imgCol + pad - 1) { // case: bottom right
paddedMat[i] = img[coordinatesToIndex(bottomRow, rightCol, imagesStrides)]; // bottom right from original img
}
else { // case: bottom, not corner
paddedMat[i] = img[coordinatesToIndex(bottomRow, currentCol, imagesStrides)]; // bottom of original image
}
}
else if (col < pad) { // case: left
paddedMat[i] = img[coordinatesToIndex(currentRow, leftCol, imagesStrides)]; // left of original image
}
else if (col > imgCol + pad -1) { // case: right
paddedMat[i] = img[coordinatesToIndex(currentRow, rightCol, imagesStrides)]; // right of original image
}
else { // case: center
paddedMat[i] = img[coordinatesToIndex(currentRow, currentCol, imagesStrides)]; // original image value
}
}
return paddedMat;
}
const result = padInputImage(xValues)
console.log(`result ${result}`)
// Interpolation kernel
function interpolationKernel(x: number, a: number): number {
if (Math.abs(x) >= 0 && Math.abs(x) <= 1) {
return (a + 2) * (Math.abs(x) ** 3) - (a + 3) * (Math.abs(x) ** 2) + 1;
}
else if (Math.abs(x) > 1 && Math.abs(x) <= 2) {
return a * (Math.abs(x)**3)-(5*a)*(Math.abs(x)**2)+(8*a)*Math.abs(x)-4*a;
}
return 0;
}
console.log(`interpolationKernel ${interpolationKernel(1,2)}`)
return backend.makeTensorInfo(
[batch, newHeight, newWidth, numChannels], 'float32', result);
}
export const resizeBicubicConfig: KernelConfig = {
kernelName: ResizeBicubic,
backendName: 'cpu',
kernelFunc: resizeBicubic as unknown as KernelFunc
};