D-lib Object Detector Training

Question

I am trying to train an object detector using D-lib. I selected about 100 images for training. I am using Python environment. According to the documentation, I used the Imglab tool to draw bounding rectangles from images. Each image is almost 4000 * 3000 pixels in size. Then he placed the generated XML file in my location and named the detector program. Here are my doubts and questions.

What should I use as a test XML file during program startup? I ran without assigning any test XML in place. However, I get the following. Learning using C: 5 Learning using epsilon: 0.01 Learning using 8 threads. Learning with a sliding window 81 pixels wide up to 79 pixels. Teaching left and right image options. slain What does "slain" mean?

What should I do now? Please guide!

Answer 1

The XML test file uses images and annotations of objects to verify the accuracy of detection of your object. The file can be generated using imglab, as with training XML. It should contain data similar to, but not identical to your training dataset.

Killed message is related to memory issues. On Linux, this is the result of OOM Killer. Since the images are so large, your machine has run out of memory and killed the learning process.
Possible solutions:
1) Set up the OOM killer to allow dlib to use more memory.
2) Reduce images if possible for your application.
3) Use a machine with a large amount of RAM. If you don't have one, AWS offers several EC2 options at a relatively low cost.