#include "QueryEngine/heavydbTypes.h"

Include dependency graph for TorchTableFunctions.h:

This graph shows which files directly or indirectly include this file:

Functions
template<typename PixelType , typename ColorType >
TEMPLATE_NOINLINE int32_t	tf_torch_raster_obj_detect__cpu_template (TableFunctionManager &mgr, const Column< PixelType > &input_x, const Column< PixelType > &input_y, const ColumnList< ColorType > &input_channels, const PixelType x_input_units_per_pixel, const PixelType y_input_units_per_pixel, const float max_color_value, const int64_t tile_boundary_halo_pixels, const TextEncodingNone &model_path, const TextEncodingNone &model_metadata_path, const float min_confidence_threshold, const float iou_threshold, const bool use_gpu, const int64_t device_num, Column< TextEncodingDict > &detected_class_label, Column< int32_t > &detected_class_id, Column< double > &detected_centroid_x, Column< double > &detected_centroid_y, Column< double > &detected_width, Column< double > &detected_height, Column< float > &detected_confidence)

Function Documentation

template<typename PixelType , typename ColorType >

TEMPLATE_NOINLINE int32_t tf_torch_raster_obj_detect__cpu_template	(	TableFunctionManager &	mgr,
		const Column< PixelType > &	input_x,
		const Column< PixelType > &	input_y,
		const ColumnList< ColorType > &	input_channels,
		const PixelType	x_input_units_per_pixel,
		const PixelType	y_input_units_per_pixel,
		const float	max_color_value,
		const int64_t	tile_boundary_halo_pixels,
		const TextEncodingNone &	model_path,
		const TextEncodingNone &	model_metadata_path,
		const float	min_confidence_threshold,
		const float	iou_threshold,
		const bool	use_gpu,
		const int64_t	device_num,
		Column< TextEncodingDict > &	detected_class_label,
		Column< int32_t > &	detected_class_id,
		Column< double > &	detected_centroid_x,
		Column< double > &	detected_centroid_y,
		Column< double > &	detected_width,
		Column< double > &	detected_height,
		Column< float > &	detected_confidence
	)

Definition at line 28 of file TorchTableFunctions.cpp.

References class_idx, detect_objects_in_tiled_raster(), get_model_info_from_file(), StringDictionaryProxy::getOrAddTransientBulk(), TextEncodingNone::getString(), TableFunctionManager::set_output_row_size(), Column< TextEncodingDict >::setNull(), and Column< TextEncodingDict >::string_dict_proxy_.

                                                                              {
   std::shared_ptr<CpuTimer> timer =
       std::make_shared<CpuTimer>("tf_torch_raster_obj_detect");
 
   timer->start_event_timer("get_model_info_from_file");
   const std::string model_path_str(model_path.getString());
   std::string model_metadata_path_str(model_metadata_path.getString());
   if (model_metadata_path_str.empty()) {
     model_metadata_path_str = model_path_str;
   }
   const auto model_info = get_model_info_from_file(model_metadata_path_str);
   if (!model_info.is_valid) {
     return mgr.ERROR_MESSAGE("Could not get model info from file.");
   }
   if (model_info.class_labels.empty()) {
     return mgr.ERROR_MESSAGE("Could not get class labels from file.");
   }
   const auto class_idx_to_label_vec =
       detected_class_label.string_dict_proxy_->getOrAddTransientBulk(
           model_info.class_labels);
   const int64_t num_class_labels = static_cast<int64_t>(class_idx_to_label_vec.size());
 
   constexpr int64_t target_batch_size_multiple{8};
 
   auto raster_data =
       RasterFormat_Namespace::format_raster_data<PixelType, ColorType, float>(
           input_x,
           input_y,
           input_channels,
           x_input_units_per_pixel,
           y_input_units_per_pixel,
           max_color_value,
           model_info.raster_tile_width,
           model_info.raster_tile_height,
           tile_boundary_halo_pixels,
           target_batch_size_multiple,
           timer->start_nested_event_timer("format_raster_data"));
 
   try {
     const auto processed_detections = detect_objects_in_tiled_raster(
         model_path_str,
         model_info,
         use_gpu,
         device_num,
         raster_data.first,
         raster_data.second,
         min_confidence_threshold,
         iou_threshold,
         timer->start_nested_event_timer("detect_objects_in_tiled_raster"));
     timer->start_event_timer("Write results");
     const int64_t num_detections = processed_detections.size();
     mgr.set_output_row_size(num_detections);
     // The class labels taken from the model file will be in same order as class idxs by
     // definition
     for (int64_t detection_idx = 0; detection_idx < num_detections; ++detection_idx) {
       const auto class_idx = processed_detections[detection_idx].class_idx;
       detected_class_id[detection_idx] = class_idx;
       if (class_idx < 0 || class_idx >= num_class_labels) {
         detected_class_label.setNull(detection_idx);
       } else {
         detected_class_label[detection_idx] = class_idx_to_label_vec[class_idx];
       }
       detected_centroid_x[detection_idx] = processed_detections[detection_idx].centroid_x;
       detected_centroid_y[detection_idx] = processed_detections[detection_idx].centroid_y;
       detected_width[detection_idx] = processed_detections[detection_idx].width;
       detected_height[detection_idx] = processed_detections[detection_idx].height;
       detected_confidence[detection_idx] = processed_detections[detection_idx].confidence;
     }
     return num_detections;
   } catch (const std::runtime_error& e) {
     return mgr.ERROR_MESSAGE(e.what());
   }
   return 0;
 }

Here is the call graph for this function: