I have an opencv code and it needs to run very fast, faster than all loops in opencv. That’s why I added most of the processes to the database, but when everything is very fast, accessing the view from the cursor is slow, according to the opencv codes.
I want to reach the view from the cursor faster. Is there any other way around this? For example, can I write a function or something so that mongocxx returns direct view to me?
I did the following operations. I share the operations with their elapsed time. I would be very happy if you can help.
void MongoDB::SecondPointIteration(cv::KeyPoint&keypoints, SecondIterParams& secondIterParams ){
auto start1 = std::chrono::steady_clock::now();
secondParams = secondIterParams;
mongocxx::pipeline secondIterPipeline = setSorgu();
mongocxx::cursor cursor = features.aggregate(secondIterPipeline, mongocxx::options::aggregate{});
auto end1 = std::chrono::steady_clock::now();
auto time = std::chrono::duration_cast<std::chrono::microseconds>(end1-start1).count();
std::cout<<"2.1: \t"<< time<<"ms"<<std::endl;
auto start3 = std::chrono::steady_clock::now();
for(auto&& doc : cursor) {
auto end3 = std::chrono::steady_clock::now();
auto time3 = std::chrono::duration_cast<std::chrono::microseconds>(end3-start3).count();
std::cout<<"2.2.1: \t"<< time3<<"ms"<<std::endl;
auto start2 = std::chrono::steady_clock::now();
...
keypoints.response = doc["_id"]["response"].get_double().value;
auto end2 = std::chrono::steady_clock::now();
auto time2 = std::chrono::duration_cast<std::chrono::microseconds>(end2-start2).count();
std::cout<<"2.2.2: \t"<< time2<<"ms"<<std::endl;
}
auto end3 = std::chrono::steady_clock::now();
auto time3 = std::chrono::duration_cast<std::chrono::microseconds>(end3-start3).count();
std::cout<<"2.2: \t"<< time3<<"ms"<<std::endl;
}
mongocxx::pipeline MongoDB::setSorgu(){
using namespace bsoncxx::builder::basic;
mongocxx::pipeline secondIterPipeline{};
...
secondIterPipeline.sort(make_document(kvp("_id.response", -1)));
secondIterPipeline.limit(1);
return secondIterPipeline;
}
Elapsed times are written by matching std :: cout
1: 761ms
2.1: 80ms
2.2.1: 1150ms
2.2.2: 2ms
2.2: 1158ms
The number 1 represents the time taken by opencv loops
My database contains similar data such as
[{
"keypoint": {
"response": 0
}
},{
"keypoint": {
"response": 0.0010120292427018285
}
},{
"keypoint": {
"response": 0.0017831606091931462
}
}]