# 创建 `srnet` 模块的步骤 Source: [https://docs.qualcomm.com/doc/80-70015-15BY/topic/steps-to-create-srnet-module.html](https://docs.qualcomm.com/doc/80-70015-15BY/topic/steps-to-create-srnet-module.html) 1. 创建模块头“gst-plugins-qti-oss/gst-plugin-base/gst/ml/ml-module-video-super-resolution.h” - 参考现有插件： [st-plugin-base/gst/ml/ml-module-video-super-resolution.h · imsdk.lnx.2.0.0.r2-rel · CodeLinaro / le / platform / vendor / qcom-opensource / gst-plugins-qti-oss · GitLab](https://git.codelinaro.org/clo/le/platform/vendor/qcom-opensource/gst-plugins-qti-oss/-/blob/imsdk.lnx.2.0.0.r2-rel/gst-plugin-base/gst/ml/ml-module-video-super-resolution.h?ref_type=heads) - 导入所需的 gst 头文件 - 创建 GST_API gboolean gst_ml_module_video_super_resolution_execute (GstMLModule * module, GstMLFrame * mlframe, GstVideoFrame * vframe);Copy to clipboard - 在调用后处理函数 (gst\_ml\_module\_process) 时需要使用该函数，该后处理函数在 ml-vsuperresolution-<module-name>.c 中定义 2. 创建模块"gst-plugins-mlvsuperresolution/modules/ml-vsuperresolution-srnet.c" - 参考现有插件： [gst-plugin-mlvsegmentation/modules/ml-vsegmentation-deeplab-argmax.c · imsdk.lnx.2.0.0.r1-rel · CodeLinaro / le / platform / vendor / qcom-opensource / gst-plugins-qti-oss · GitLab](https://git.codelinaro.org/clo/le/platform/vendor/qcom-opensource/gst-plugins-qti-oss/-/blob/imsdk.lnx.2.0.0.r1-rel/gst-plugin-mlvsegmentation/modules/ml-vsegmentation-deeplab-argmax.c?ref_type=heads) - 导入"modules/ml-video-superresolution-module.h"头文件 - 定义模型输出 caps #define GST_ML_MODULE_CAPS \ "neural-network/tensors, " \ "type = (string) { INT8, UINT8, INT32, FLOAT32 }, " \ "dimensions = (int) < <1, [32, 4096], [32, 4096]> >; " \ "neural-network/tensors, " \ "type = (string) { INT8, UINT8, INT32, FLOAT32 }, " \ "dimensions = (int) < <1, [32, 4096], [32, 4096], [1, 3]> >"Copy to clipboard **例如** - 从 AI Hub 下载 [QuickSRNetLarge-Quantized Qualcomm AI Hub](https://aihub.qualcomm.com/iot/models/quicksrnetlarge_quantized) 模型 - 检查 TFLite 模型的输出尺寸（参见 [在应用程序中集成 AI Hub 模型](https://docs.qualcomm.com/doc/80-70015-15BY/topic/integrate-aihub-model.html)章节） - 格式：dtype[N x H X W X C] - 在以上模块输出 caps <1, [32, 4096], [32, 4096], [1, 3]> 中 - N = 1 -> 模型输出 Batch 处理大小 - H = [32,4096] -> 模型输出高度应在此范围内 - W = [32,4096] -> 模型输出宽度应在此范围内 - C = [1,3] -> 模型输出通道数应在此范围内 - dtype = int8（量化模型）| float32（非量化模型） - 对于量化模型，检查输出节点的缩放系数和偏置 - 格式：scale\*(q-offset) - 实现以下函数 - 初始化 - `gpointer gst_ml_module_open (void);` - 反初始化 - `void gst_ml_module_close (gpointer instance);` - Caps - `GstCaps *gst_ml_module_caps (void);` - 配置 - `gboolean gst_ml_module_configure (gpointer instance, GstStructure * settings);` - 检查模块的数据格式、缩放系数和偏置 - 反量化 gdouble gst_ml_tensor_extract_value (GstMLType mltype, gpointer data, guint idx, gdouble offset, gdouble scale)Copy to clipboard - 需要传递给量化模型的缩放系数和偏置值 - 后处理 - `gboolean gst_ml_module_process (gpointer instance, GstMLFrame*mlframe, gpointer output);` - 推理缓冲区 - `indata = GST_ML_FRAME_BLOCK_DATA (mlframe, 0);` 用后处理数据填充的输出缓冲区 - `outdata = GST_VIDEO_FRAME_PLANE_DATA (vframe, 0);` 模型的数据类型 - `mltype = GST_ML_FRAME_TYPE (mlframe);` 后处理（用于超分辨率） for (row = 0; row < GST_VIDEO_FRAME_HEIGHT (vframe); row++) { for (column = 0; column < GST_VIDEO_FRAME_WIDTH (vframe); column++) { // Calculate the destination index. idx = (((row * GST_VIDEO_FRAME_WIDTH (vframe)) + column) * bpp) + (row * padding); outdata[idx] = gst_ml_tensor_extract_value (mltype, indata, idx, submodule->qoffsets[0], submodule->qscales[0]); outdata[idx + 1] = gst_ml_tensor_extract_value (mltype, indata, (idx + 1), submodule->qoffsets[0], submodule->qscales[0]); outdata[idx + 2] = gst_ml_tensor_extract_value (mltype, indata, (idx + 2), submodule->qoffsets[0], submodule->qscales[0]); // If output has an alpha channel set it to opaque. if (bpp == 4) outdata[idx + 3] = 0xFF; } }Copy to clipboard - 更新"gst-plugins-mlvsuperresolution/modules/CMakeLists.txt"中的模块名称 - 来自现有插件的参考： [gst-plugin-mlvsegmentation/modules/CMakeLists.txt · imsdk.lnx.2.0.0.r1-rel · CodeLinaro / le / platform / vendor / qcom-opensource / gst-plugins-qti-oss · GitLab](https://git.codelinaro.org/clo/le/platform/vendor/qcom-opensource/gst-plugins-qti-oss/-/blob/imsdk.lnx.2.0.0.r1-rel/gst-plugin-mlvsegmentation/modules/CMakeLists.txt?ref_type=heads) Last Published: Jan 26, 2026 [Previous Topic 模块概述](https://docs.qualcomm.com/bundle/publicresource/80-70015-15BY/topics/overview-of-module.md) [Next Topic 更新 BitBake 文件以进行插件编译的步骤](https://docs.qualcomm.com/bundle/publicresource/80-70015-15BY/topics/steps-to-update-bitbake-for-compilation.md)