Function main

tests/test_sam2_video_debug.cpp:51–176

}

int main(int argc, char** argv) {
    if (argc < 3) {
        fprintf(stderr, "Usage: %s <model.ggml> <video_frames_dir> [point_x point_y]\n", argv[0]);
        return 1;
    }

    const char* model_path = argv[1];
    const char* frames_dir = argv[2];
    float point_x = (argc > 3) ? atof(argv[3]) : 210.0f;
    float point_y = (argc > 4) ? atof(argv[4]) : 350.0f;

    const char* dump_dir = "/tmp/debug_sam2_cpp";
    mkdir(dump_dir, 0755);

    const int N_FRAMES = argc > 5 ? atoi(argv[5]) : 5;
    int encode_img_size = (argc > 6) ? atoi(argv[6]) : 0;

    // Load model
    fprintf(stderr, "Loading model: %s\n", model_path);
    sam3_params params;
    params.model_path = model_path;
    params.use_gpu = false;  // CPU for reproducibility
    params.encode_img_size = encode_img_size;
    auto model = sam3_load_model(params);
    if (!model) {
        fprintf(stderr, "Failed to load model\n");
        return 1;
    }

    // Create state
    auto state = sam3_create_state(*model, params);
    if (!state) {
        fprintf(stderr, "Failed to create state\n");
        return 1;
    }

    // Load frames as JPEG images
    std::vector<sam3_image> frames;
    for (int i = 0; i < N_FRAMES + 1; ++i) {
        char path[512];
        snprintf(path, sizeof(path), "%s/%05d.jpg", frames_dir, i);
        sam3_image img = sam3_load_image(path);
        if (img.data.empty()) {
            fprintf(stderr, "Failed to load frame %d: %s\n", i, path);
            return 1;
        }
        fprintf(stderr, "Loaded frame %d: %dx%d\n", i, img.width, img.height);
        frames.push_back(std::move(img));
    }

    // Create visual tracker
    sam3_visual_track_params vtp;
    vtp.max_keep_alive = 100;
    vtp.recondition_every = 16;
    auto tracker = sam3_create_visual_tracker(*model, vtp);
    if (!tracker) {
        fprintf(stderr, "Failed to create tracker\n");
        return 1;