NFeature | Simple feature configuration and availability-control | Frontend Framework library

I am working on a machine learning problem, where I have a lot of zipcodes (~8k unique values) in my data set. Thus I decided to hash the values into a smaller feature space instead of using something like OHE.

The problem I encountered was a very small percentage (20%) of unique rows in my hash, which basically means from my understanding, that I have a lot of duplicates/collisions. Even though I increased the features in my hash table to ~200, I never got more than 20% of unique values. This does not make sense to me, since with a growing number of columns in my hash, more unique combinations should be possible

I used the following code to hash my zip codes with scikit and calculate the collisions based on unique vales in the last array:

I'm looking at the MediaEval 2018 memorability challenge (here).

One of the features they describe is ORB features. I got the data from the challenge, and I'm trying to understand how the ORB data works.

If I run this code:

I want to get mapping to work using a Picamera. I have a Raspberry Pi running a cv_camera_node and an Ubuntu 20.04.1 running roscore, as well as, slam and rviz. I have OpenCV 4.2.0 and installed the following version of orb-slam2: https://github.com/appliedAI-Initiative/orb_slam_2_ros. I am running ROS Noetic. I have wrote the following launch file for slam:

My full code:

I am using a MLP model for classification.

When I predict for new data, I want to keep only those predictions whose probability of prediction is larger than 0.5, and change all other predictions into class 0.

How can I do it in keras ?

I'm using using last layer as follows model.add(layers.Dense(7 , activation='softmax'))

Is it meaningful to get predictions with probability larger than 0.5 using the softmax?

Consider a plotly figure where you can select polynomial features for a line fit using JupyterDash:

If you select an area and then choose another number for polynomial features, the figure goes from this:

... and back to this again:

So, how can you set things up so that the figure displays the same area of the figure every time you select another number of features and trigger another callback?

When running grid search on Inverse of regularization strength parameter and number of nearest neighbors parameter for logistic regression , linear SVM and K nearest neighbors classifier , The best parameters obtained from gridsearch are not really the best when verifying manually by training on same training data set. Code below

ORB doesn't find keypoints near the edge of an image and I don't understand why. It seems worse that SIFT and SURF and I would expect the opposite.

If I understand correctly then SIFT/SURF use a 16x16 and 20x20 square block respectedly around the test-point so I would expect them not to find keypoints 8 and 10 pixels from an edge. FAST/ORB uses a circle of diameter 7 around the test-point so I expected it to find keypoints even closer to the edge, perhaps as close as 4 pixels (though I think the associated algorithm, BRIEF, to describe keypoints uses a larger window so this would remove some keypoints).

An experiment makes nonsense of my prediction. The minimum distance from the edge in my experiments vary with the size and spacing of the squares but examples are

• SIFT .. 5 pixels
• SURF .. 15 pixels
• ORB .. 39 pixels

Can anyone explain why?

The code I used is below. I drew a grid of squares and applied a Gaussian blur. I expected the algorithms to latch onto the corners but they found the centres of the squares and some artifacts.