wip: C++ demo 支持批量清单

deploy/cpp/main.cpp

@@ -1,9 +1,12 @@
 // 演示与对拍工具：读一张原始像素转储，推理并打印框。
 //
-//   ./film_predict <model.xml> <image.bin> <宽> <高> <通道> [conf] [tile]
+//   单张: ./film_predict <model.xml> <image.bin> <宽> <高> <通道> [conf] [overlap]
+//   批量: ./film_predict <model.xml> --manifest <清单> [conf] [overlap]
 //
 // 输入是连续的 HWC uint8 缓冲（用 scripts/dump_raw.py 从图片生成）。产线里通常直接
-// 拿相机帧的指针，不需要这一步。tile 传 0 表示图本身就是网络输入尺寸。
+// 拿相机帧的指针，不需要这一步。
+//
+// 清单每行 `路径 宽 高 通道`，模型只加载一次——逐张起进程的话加载耗时会盖过推理。
 #include "film_detector.hpp"
 
 #include <cstdio>
@@ -11,39 +14,70 @@
 #include <iostream>
 #include <vector>
 
-int main(int argc, char** argv) {
+namespace {
-    if (argc < 6) {
-        std::cerr << "用法: " << argv[0]
-                  << " <model.xml> <image.bin> <宽> <高> <通道> [conf] [tile-overlap]\n";
-        return 1;
-    }
-    const std::string model_path = argv[1], image_path = argv[2];
-    const int width = std::stoi(argv[3]), height = std::stoi(argv[4]);
-    const int channels = std::stoi(argv[5]);
-    const float conf = argc > 6 ? std::stof(argv[6]) : 0.25f;
-    const int overlap = argc > 7 ? std::stoi(argv[7]) : 32;
 
-    std::ifstream file(image_path, std::ios::binary);
+bool read_pixels(const std::string& path, size_t expected, std::vector<unsigned char>& out) {
-    if (!file) { std::cerr << "打不开 " << image_path << "\n"; return 1; }
+    std::ifstream file(path, std::ios::binary);
-    std::vector<unsigned char> pixels((std::istreambuf_iterator<char>(file)),
+    if (!file) { std::cerr << "打不开 " << path << "\n"; return false; }
-                                       std::istreambuf_iterator<char>());
+    out.assign(std::istreambuf_iterator<char>(file), std::istreambuf_iterator<char>());
-    const size_t expected = static_cast<size_t>(width) * height * channels;
+    if (out.size() != expected) {
-    if (pixels.size() != expected) {
+        std::cerr << path << " 的缓冲是 " << out.size() << " 字节，应为 " << expected << "\n";
-        std::cerr << "缓冲大小 " << pixels.size() << " 与 " << width << "x" << height
+        return false;
-                  << "x" << channels << " 不符（应为 " << expected << "）\n";
-        return 1;
     }
+    return true;
+}
 
-    film::Detector detector(model_path);
+void run_one(const film::Detector& detector, const film::ImageView& image, float conf,
-    film::ImageView image{pixels.data(), width, height, channels};
+             int overlap) {
-    const auto boxes = (width == detector.input_size() && height == detector.input_size())
+    const auto boxes = (image.width == detector.input_size() &&
+                        image.height == detector.input_size())
         ? detector.infer_tile(image, conf)
         : detector.infer(image, conf, 0.5f, overlap);
-
     std::printf("%zu\n", boxes.size());
     for (const auto& b : boxes) {
         std::printf("%d %.6f %.6f %.6f %.6f %.4f\n",
                     b.class_id, b.x1, b.y1, b.x2, b.y2, b.score);
     }
+}
+
+}  // namespace
+
+int main(int argc, char** argv) {
+    if (argc < 4) {
+        std::cerr << "用法:\n  " << argv[0] << " <model.xml> <image.bin> <宽> <高> <通道>"
+                  << " [conf] [overlap]\n  " << argv[0]
+                  << " <model.xml> --manifest <清单> [conf] [overlap]\n";
+        return 1;
+    }
+    const std::string model_path = argv[1];
+    const bool batch = std::string(argv[2]) == "--manifest";
+    const float conf = batch ? (argc > 4 ? std::stof(argv[4]) : 0.25f)
+                             : (argc > 6 ? std::stof(argv[6]) : 0.25f);
+    const int overlap = batch ? (argc > 5 ? std::stoi(argv[5]) : 32)
+                              : (argc > 7 ? std::stoi(argv[7]) : 32);
+
+    film::Detector detector(model_path);
+    std::vector<unsigned char> pixels;
+
+    if (batch) {
+        std::ifstream manifest(argv[3]);
+        if (!manifest) { std::cerr << "打不开清单 " << argv[3] << "\n"; return 1; }
+        std::string path;
+        int width = 0, height = 0, channels = 0;
+        while (manifest >> path >> width >> height >> channels) {
+            if (!read_pixels(path, static_cast<size_t>(width) * height * channels, pixels)) {
+                return 1;
+            }
+            run_one(detector, film::ImageView{pixels.data(), width, height, channels},
+                    conf, overlap);
+        }
+        return 0;
+    }
+
+    if (argc < 6) { std::cerr << "单张模式需要 宽 高 通道\n"; return 1; }
+    const int width = std::stoi(argv[3]), height = std::stoi(argv[4]);
+    const int channels = std::stoi(argv[5]);
+    if (!read_pixels(argv[2], static_cast<size_t>(width) * height * channels, pixels)) return 1;
+    run_one(detector, film::ImageView{pixels.data(), width, height, channels}, conf, overlap);
     return 0;
 }