Add Bounding Box Crop Function

CMakeLists.txt

@@ -26,6 +26,7 @@ set(CMAKE_MODULE_LINKER_FLAGS "${CMAKE_MODULE_LINKER_FLAGS} -fuse-ld=lld")
 set(SOURCES
     "${FusswegDatentools_SOURCE_DIR}/src/main.cpp"
     "${FusswegDatentools_SOURCE_DIR}/src/annot.cpp"
+    "${FusswegDatentools_SOURCE_DIR}/src/img.cpp"
     "${FusswegDatentools_SOURCE_DIR}/src/via.cpp"
     "${FusswegDatentools_SOURCE_DIR}/src/crs.cpp"
     "${FusswegDatentools_SOURCE_DIR}/src/cv.cpp"
@@ -34,6 +35,7 @@ set(SOURCES
 )
 set(HEADERS
     "${FusswegDatentools_SOURCE_DIR}/include/annot.hpp"
+    "${FusswegDatentools_SOURCE_DIR}/include/img.hpp"
     "${FusswegDatentools_SOURCE_DIR}/include/via.hpp"
     "${FusswegDatentools_SOURCE_DIR}/include/crs.hpp"
     "${FusswegDatentools_SOURCE_DIR}/include/cv.hpp"

include/img.hpp

@@ -0,0 +1,14 @@
+#pragma once
+
+#include <string>
+
+namespace fdt {
+
+    namespace img {
+
+        void bboxCrop(const std::string &, const std::string &,
+                      const std::string &, const int, const int);
+
+    } // namespace img
+
+} // namespace fdt

include/utils.hpp

@@ -7,6 +7,7 @@
 #include <vector>
 
 #define MAX2(x, y) ((x) > (y) ? (x) : (y))
+#define MIN2(x, y) ((x) < (y) ? (x) : (y))
 
 typedef std::vector<std::string> Paths;
 

src/img.cpp

@@ -0,0 +1,81 @@
+#include "img.hpp"
+#include "csv.hpp"
+#include "utils.hpp"
+#include <future>
+#include <iostream>
+#include <opencv2/opencv.hpp>
+
+// Function to process each record and save the bounding box as an image
+// Handles one TSV file at a time
+void crop_bbox(const std::string &root_dir, const std::string &tsv_file,
+               const std::string &output_dir, const int width,
+               const int height) {
+    // Create a TSV reader
+    csv::CSVFormat format;
+    format.delimiter('\t').header_row(0);
+    csv::CSVReader reader(tsv_file, format);
+
+    // Iterate over each row
+    for (auto &row : reader) {
+        const auto prefix = row["prefix"].get<std::string>();
+        const auto image_name = row["image"].get<std::string>();
+        const auto category = row["cate"].get<std::string>();
+        const auto level = row["level"].get<std::string>();
+
+        // Assertion:
+        // - x = max(0, x); y = max(0, y)
+        // - w = min(w, width - x); h = min(h, height - y)
+        const int x = MAX2(0, row["x"].get<int>());
+        const int y = MAX2(0, row["y"].get<int>());
+        const int w = MIN2(row["w"].get<int>(), width - x);
+        const int h = MIN2(row["h"].get<int>(), height - y);
+
+        // Load the image with std::filesystem by combining the root dir,
+        // prefix, and image name
+        std::string image_path =
+            std::filesystem::path(root_dir) / prefix / image_name;
+
+        cv::Mat image = cv::imread(image_path);
+        // NOTE: cv::imread() silently returns an empty matrix if it fails to
+        // load the image, rendering try-catch blocks useless
+        if (image.empty()) {
+            std::cerr << "Could not open or find the image: " << image_path
+                      << std::endl;
+            continue;
+        }
+
+        // Extract the bounding box
+        cv::Rect bounding_box(x, y, w, h);
+        cv::Mat cropped_image = image(bounding_box);
+
+        // Create the output image name
+        std::string output_name =
+            "_" + category + "_" + level + "_" + prefix + "_" + image_name;
+
+        // Save the cropped image
+        std::string output_path = output_dir + "/" + output_name;
+        imwrite(output_path, cropped_image);
+    }
+}
+
+// Crop the bounding box and save the image.
+// TSV files in `tsv_dir` will all be read first and combined, before being
+// shuffled and split to different threads, each of which will then process
+// the images and save them to the output directory.
+void fdt::img::bboxCrop(const std::string &root_dir, const std::string &tsv_dir,
+                        const std::string &output_path, const int width,
+                        const int height) {
+    // Vector to hold future objects
+    std::vector<std::future<void>> futures;
+
+    // Launch a thread for each TSV file
+    for (const auto &f : fdt::utils::listAllFiles(tsv_dir, ".tsv")) {
+        futures.push_back(std::async(std::launch::async, crop_bbox, root_dir, f,
+                                     output_path, width, height));
+    }
+
+    // Wait for all threads to finish
+    for (auto &f : futures) {
+        f.get();
+    }
+}

src/main.cpp

@@ -6,6 +6,7 @@
 #include "cv.hpp"
 #include "exif.hpp"
 #include "gis.hpp"
+#include "img.hpp"
 #include "utils.hpp"
 #include "via.hpp"
 
@@ -30,6 +31,9 @@ int parse_args(int argc, char *argv[]) {
                   << "<label_dir>" << std::endl;
         std::cout << "  " << argv[0] << " annot-to-coco "
                   << "<annot_dir> <exif_dir> <output_file>" << std::endl;
+        std::cout << "  " << argv[0] << " crop-bbox "
+                  << "<root_dir> <tsv_dir> <output_dir> <width> <height>"
+                  << std::endl;
         std::cout << "  " << argv[0] << " bbox-draw "
                   << "<label_dir> <src_dir> <dst_dir> <ext>" << std::endl;
         std::cout << "  " << argv[0] << " crs-to-nzgd2000 "
@@ -55,9 +59,9 @@ int parse_args(int argc, char *argv[]) {
     if (op != "exif-export-json" && op != "exif-export-csv" &&
         op != "displacement" && op != "via-export" && op != "via-print-stats" &&
         op != "via-export-stats" && op != "annot-to-coco" &&
-        op != "bbox-draw" && op != "pov-roi" && op != "pov-transform" &&
-        op != "crs-to-nzgd2000" && op != "crs-from-nzgd2000" &&
-        op != "geojson-to-tsv") {
+        op != "crop-bbox" && op != "bbox-draw" && op != "pov-roi" &&
+        op != "pov-transform" && op != "crs-to-nzgd2000" &&
+        op != "crs-from-nzgd2000" && op != "geojson-to-tsv") {
         throw std::runtime_error("Unknown operation. ");
     }
     if ((op == "exif-export-json" && argc != 4) ||
@@ -67,7 +71,7 @@ int parse_args(int argc, char *argv[]) {
         (op == "via-export-stats" && argc != 4) ||
         (op == "via-print-stats" && argc != 3) ||
         (op == "annot-to-coco" && argc != 5) ||
-        (op == "bbox-draw" && argc != 6) ||
+        (op == "crop-bbox" && argc != 7) || (op == "bbox-draw" && argc != 6) ||
         (op == "geojson-to-tsv" && argc != 4) ||
         (op == "crs-to-nzgd2000" && argc != 4) ||
         (op == "crs-from-nzgd2000" && argc != 4) ||
@@ -132,6 +136,16 @@ int parse_args(int argc, char *argv[]) {
         fdt::via::printStats(dir_lab);
         return 0;
     }
+    if (op == "crop-bbox") {
+        std::string root_dir = argv[2];
+        std::string tsv_dir = argv[3];
+        std::string out_dir = argv[4];
+        int width = std::strtol(argv[5], nullptr, 10);
+        int height = std::strtol(argv[6], nullptr, 10);
+        fdt::img::bboxCrop(root_dir, tsv_dir, out_dir, width, height);
+        return 0;
+    }
+
     if (op == "annot-to-coco") {
         std::string dir_annot = argv[2];
         std::string dir_exif = argv[3];