From the depths to the surface,
defaultUser
OverviewAs some readers may know, I have a project that is in the works called Project Lädra. This project's mission is to take a video from a camera, 2D or 3D, to identify cleanable surfaces. This is accomplished by using a neural network (imagine different neurons in your brain) to identify different objects in an image, taken from the video. Current ProgressFor the project up until this point, it has been going according to plan, with some fall back. As of now, I have the tools to put together a network together, except one: an image dataset. The only tool needed is because this network needs to be trained on, just like the neurons in your brain need to be trained how to walk and such. Grabbing a dataset is not as easy as it sounds. The ImageNet Dataset that I am going to use is only given via request or URL grabbing (taking an image's web location and copying the image to your storage) Other than that hurdle there is not that many kinks to iron out to this project. The project is eclipsing on the experimental phase, where model tests are going to be run. ConclusionProject Lädra is a bit behind course due to a hurdle in this project. After that hurdle is traversed, an experimental process can be started and the project can return to its original path of modifying variables to improve on the network.
0 Comments
Leave a Reply. |
Author:
|