[SYSTEMDS-3546] Push down image pre-processing to blas/mkl #1843
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,243 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one | ||
| * or more contributor license agreements. See the NOTICE file | ||
| * distributed with this work for additional information | ||
| * regarding copyright ownership. The ASF licenses this file | ||
| * to you 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. | ||
| */ | ||
|
|
||
| #include <iostream> | ||
| #include <cmath> | ||
| #include <cstring> | ||
| #include "common.h" | ||
|
|
||
| using namespace std; | ||
|
|
||
| void printImage(const double* image, int rows, int cols) { | ||
| for (int i = 0; i < rows; i++) { | ||
| for (int j = 0; j < cols; j++) { | ||
| cout << image[i * cols + j] << " "; | ||
| } | ||
| cout << endl; | ||
| } | ||
| cout << "\n"<< endl; | ||
| } | ||
|
There was a problem hiding this comment. i suggest removing this debugging printImage. |
||
|
|
||
| void img_transform(const double* img_in, int orig_w, int orig_h, int out_w, int out_h, double a, double b, double c, double d, | ||
| double e, double f, double fill_value, double* img_out) { | ||
| double divisor = a * e - b * d; | ||
| if (divisor == 0) { | ||
| std::cout << "Inverse matrix does not exist! Returning input." << std::endl; | ||
| for (int i = 0; i < orig_h; i++) { | ||
| for (int j = 0; j < orig_w; j++) { | ||
| img_out[i * orig_w + j] = img_in[i * orig_w + j]; | ||
| } | ||
| } | ||
| } else { | ||
| // Create the inverted transformation matrix | ||
| double T_inv[9]; | ||
| T_inv[0] = e / divisor; | ||
| T_inv[1] = -b / divisor; | ||
| T_inv[2] = (b * f - c * e) / divisor; | ||
| T_inv[3] = -d / divisor; | ||
| T_inv[4] = a / divisor; | ||
| T_inv[5] = (c * d - a * f) / divisor; | ||
| T_inv[6] = 0.0; | ||
| T_inv[7] = 0.0; | ||
| T_inv[8] = 1.0; | ||
|
|
||
| // Create the coordinates of output pixel centers linearized in row-major order | ||
| double* coords = new double[2 * out_w * out_h]; | ||
| for (int i = 0; i < out_h; i++) { | ||
| for (int j = 0; j < out_w; j++) { | ||
| coords[2 * (i * out_w + j)] = j + 0.5; | ||
| coords[2 * (i * out_w + j) + 1] = i + 0.5; | ||
| } | ||
| } | ||
|
|
||
| // Perform matrix multiplication to compute sampling pixel indices | ||
| double* transformed_coords = new double[2 * out_w * out_h]; | ||
|
There was a problem hiding this comment. i do not know, but can this be allocated as int? |
||
| for (int i = 0; i < out_w * out_h; i++) { | ||
| double x = coords[2 * i]; | ||
| double y = coords[2 * i + 1]; | ||
| transformed_coords[2 * i] = std::floor(T_inv[0] * x + T_inv[1] * y + T_inv[2]) + 1; | ||
| transformed_coords[2 * i + 1] = std::floor(T_inv[3] * x + T_inv[4] * y + T_inv[5]) + 1; | ||
| } | ||
|
|
||
| // Fill output image | ||
| for (int i = 0; i < out_h; i++) { | ||
| for (int j = 0; j < out_w; j++) { | ||
| int inx = static_cast<int>(transformed_coords[2 * (i * out_w + j)]) - 1; | ||
| int iny = static_cast<int>(transformed_coords[2 * (i * out_w + j) + 1]) - 1; | ||
| if (inx >= 0 && inx < orig_w && iny >= 0 && iny < orig_h) { | ||
| img_out[i * out_w + j] = img_in[iny * orig_w + inx]; | ||
| } else { | ||
| img_out[i * out_w + j] = fill_value; | ||
| } | ||
| } | ||
| } | ||
|
|
||
| delete[] coords; | ||
| delete[] transformed_coords; | ||
| } | ||
| } | ||
|
|
||
| void imageRotate(double* img_in, int rows, int cols, double radians, double fill_value, double* img_out) { | ||
|
|
||
| // Translation matrix for moving the origin to the center of the image | ||
| double t1_data[] = { | ||
| 1, 0, static_cast<double>(-cols)/2, | ||
| 0, 1, static_cast<double>(-rows)/2, | ||
| 0, 0, 1 | ||
| }; | ||
| double* t1 = t1_data; | ||
| // Translation matrix for moving the origin back to the top left corner | ||
| double t2_data[] = { | ||
| 1, 0, static_cast<double>(cols)/2, | ||
| 0, 1, static_cast<double>(rows)/2, | ||
| 0, 0, 1 | ||
| }; | ||
| double* t2 = t2_data; | ||
| // The rotation matrix around the origin | ||
| double rot_data[] = { | ||
| cos(radians), sin(radians), 0, | ||
| -sin(radians), cos(radians), 0, | ||
| 0, 0, 1 | ||
| }; | ||
| double* rot = rot_data; | ||
|
|
||
| // Combined transformation matrix | ||
| double m_data1[3*3]; | ||
| cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, 3, 3, 3, 1.0, t2, 3, rot, 3, 0.0, m_data1, 3); | ||
| double m_data2[3*3]; | ||
| cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, 3, 3, 3, 1.0, m_data1, 3, t1, 3, 0.0, m_data2, 3); | ||
| double* m = m_data2; | ||
| // Transform image | ||
| img_transform(img_in,rows,cols,rows,cols,m[0], m[1], m[2], m[3], m[4], m[5], fill_value, img_out); | ||
| } | ||
|
|
||
| double* imageCutout(double* img_in, int rows, int cols, int x, int y, int width, int height, double fill_value) { | ||
| // Allocate memory for the output image using MKL | ||
| double* img_out = new double[rows * cols]; | ||
|
|
||
| if (width < 1 || height < 1) { | ||
| // Invalid width or height, return the input image as it is | ||
| cblas_dcopy(rows * cols, img_in, 1, img_out, 1); | ||
| } else { | ||
| int end_x = x + width - 1; | ||
| int end_y = y + height - 1; | ||
|
|
||
| int start_x = std::max(1, x); | ||
| int start_y = std::max(1, y); | ||
| end_x = std::min(cols, end_x); | ||
| end_y = std::min(rows, end_y); | ||
|
|
||
| // Copy the input image to the output image using MKL | ||
| cblas_dcopy(rows * cols, img_in, 1, img_out, 1); | ||
|
|
||
| // Fill the cutout region with the fill_value | ||
| for (int i = start_y - 1; i < end_y; ++i) { | ||
| for (int j = start_x - 1; j < end_x; ++j) { | ||
| img_out[i * cols + j] = fill_value; | ||
| } | ||
| } | ||
| } | ||
|
|
||
| return img_out; | ||
| } | ||
|
|
||
| double* imageCrop(double* img_in, int orig_w, int orig_h, int w, int h, int x_offset, int y_offset) { | ||
| // Allocate memory for the output image | ||
| double* img_out = new double[w * h]; | ||
|
|
||
| int start_h = (std::ceil((orig_h - h) / 2)) + y_offset - 1 ; | ||
| int end_h = (start_h + h - 1); | ||
| int start_w = (std::ceil((orig_w - w) / 2)) + x_offset - 1; | ||
| int end_w = (start_w + w - 1); | ||
|
|
||
| // Create a mask to identify the cropped region | ||
| double* mask = new double[orig_w * orig_h]; | ||
| double* temp_mask = new double[w * h]; | ||
|
|
||
| // Set mask elements to 0 outside the cropped region and 1 inside | ||
| memset(mask, 0, orig_w * orig_h * sizeof(double)); | ||
| for(int i = 0; i < h * w; i++) { | ||
| temp_mask[i] = 1; | ||
| } | ||
|
|
||
| for (int i = start_h; i <= end_h; ++i) { | ||
| for (int j = start_w; j <= end_w; ++j) { | ||
| mask[i * orig_w + j] = temp_mask[(i - start_h) * w + (j - start_w)]; | ||
| } | ||
| } | ||
|
|
||
| // Apply the mask to crop the image | ||
| for (int i = 0; i < h; ++i) { | ||
| for (int j = 0; j < w; ++j) { | ||
| img_out[i * w + j] = img_in[(start_h + i) * orig_w + (start_w + j)] * mask[(start_h + i) * orig_w + (start_w + j)]; | ||
| } | ||
| } | ||
|
|
||
| // Free memory for the mask | ||
| delete[] mask; | ||
| delete[] temp_mask; | ||
|
|
||
| return img_out; | ||
| } | ||
|
|
||
| void img_translate(double* img_in, double offset_x, double offset_y, | ||
|
There was a problem hiding this comment. cast the offsets in the first line rather than every call to offset_x and y, and maybe consider changing the API to use int offsets. |
||
| int in_w, int in_h, int out_w, int out_h, double fill_value, double* img_out) { | ||
| int w = out_w; | ||
| int h = out_h; | ||
|
|
||
| offset_x = round(offset_x); | ||
| offset_y = round(offset_y); | ||
|
|
||
|
|
||
| int start_x = 0 - static_cast<int>(offset_x); | ||
| int start_y = 0 - static_cast<int>(offset_y); | ||
| int end_x = std::max(w, out_w) - static_cast<int>(offset_x); | ||
| int end_y = std::max(h, out_h) - static_cast<int>(offset_y); | ||
|
|
||
| if (start_x < 0) | ||
| start_x = 0; | ||
| if (start_y < 0) | ||
| start_y = 0; | ||
|
|
||
| if (w < end_x) | ||
| end_x = w; | ||
| if (h < end_y) | ||
| end_y = h; | ||
|
|
||
| if (out_w < end_x + static_cast<int>(offset_x)) | ||
| end_x = out_w - static_cast<int>(offset_x); | ||
| if (out_h < end_y + static_cast<int>(offset_y)) | ||
| end_y = out_h - static_cast<int>(offset_y); | ||
|
|
||
| for (int y = 0; y < out_h; ++y) { | ||
| for (int x = 0; x < out_w; ++x) { | ||
| img_out[y * out_w + x] = fill_value; | ||
| } | ||
| } | ||
|
|
||
| if (start_x < end_x && start_y < end_y) { | ||
| for (int y = start_y + static_cast<int>(offset_y); y < end_y + static_cast<int>(offset_y); ++y) { | ||
| int x_in = (start_x > 0) ? start_x + static_cast<int>(offset_x): start_x; | ||
| int y_in = (start_y > 0 ) ? y : y - static_cast<int>(offset_y); | ||
| cblas_dcopy(end_x - start_x, &img_in[(x_in) + (y_in) * in_w], | ||
| 1, &img_out[y * out_w + start_x + static_cast<int>(offset_x)], 1); | ||
| } | ||
| } | ||
|
|
||
| } | ||
|
|
||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,38 @@ | ||
| /* | ||
| * Licensed to the Apache Software Foundation (ASF) under one | ||
| * or more contributor license agreements. See the NOTICE file | ||
| * distributed with this work for additional information | ||
| * regarding copyright ownership. The ASF licenses this file | ||
| * to you 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. | ||
| */ | ||
|
|
||
| #ifndef IMAGE_UTILS_H | ||
| #define IMAGE_UTILS_H | ||
| #include <cmath> | ||
|
|
||
| void printImage(const double* image, int rows, int cols); | ||
|
|
||
|
|
||
| void m_img_transform(const double* img_in, int orig_w, int orig_h, int out_w, int out_h, double a, double b, double c, double d, | ||
| double e, double f, double fill_value, double* img_out); | ||
|
|
||
| void imageRotate(double* img_in, int rows, int cols, double radians, double fill_value, double* img_out); | ||
|
|
||
| double* imageCutout(double* img_in, int rows, int cols, int x, int y, int width, int height, double fill_value); | ||
|
|
||
| double* imageCrop(double* img_in, int orig_w, int orig_h, int w, int h, int x_offset, int y_offset); | ||
|
|
||
| void img_translate(double* img_in, double offset_x, double offset_y, | ||
| int in_w, int in_h, int out_w, int out_h, double fill_value, double* img_out); | ||
|
|
||
| #endif /* IMAGE_UTILS_H */ | ||
|
|
Binary file not shown.
Binary file not shown.
Oops, something went wrong.
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we need to revert here, since the change is only syntax.