I am running the noetic branch with ROS Noetic on Ubuntu 20.04.
I noticed that when I train and test in the same program run, I get good results for the random forest laser detector, but when I load a trained model I get garbage.
My model was trained with:
rf_num_trees 15
rf_max_depth 10
But when I load it, I see that m_randomForest->getTermCriteria().maxCount is 50, instead of 15.
I noticed that in the function RandomForestDetector::trainOnFeatures the ROS parameters are considered. But if I load a detector model from a file and then attempt to classify - those parameters are not considered.
I am now getting the ROS params in the constructor, which seems to fix this issue.
Can you look into it, and tell me if I am doing something incorrectly?
I am running the noetic branch with ROS Noetic on Ubuntu 20.04. I noticed that when I train and test in the same program run, I get good results for the random forest laser detector, but when I load a trained model I get garbage.
My model was trained with: rf_num_trees 15 rf_max_depth 10 But when I load it, I see that m_randomForest->getTermCriteria().maxCount is 50, instead of 15.
I noticed that in the function RandomForestDetector::trainOnFeatures the ROS parameters are considered. But if I load a detector model from a file and then attempt to classify - those parameters are not considered. I am now getting the ROS params in the constructor, which seems to fix this issue.
Can you look into it, and tell me if I am doing something incorrectly?