Closed JackKelly closed 2 years ago
Probably go back to using absolute encoding.
Need sufficient channels so the model can understand where the whole image is in time and space, and also which pixels are next to each other, etc.
Base this on code from my "December" model. But with tests!
This is a duplicate of issue #31
Detailed Description
Probably go back to using absolute encoding.
Need sufficient channels so the model can understand where the whole image is in time and space, and also which pixels are next to each other, etc.
Possible Implementation
Base this on code from my "December" model. But with tests!