Computer Vision News - September 2018

That’s all well and good; however, to be able to better analyze the interaction between neurons we now want to align features. We’ll do this by using the function lowres-tensor . This function gets two shape objects and produces a per-dimension bilinear interpolation between those shapes. Let’s get an idea of what this means with a simple case: Let's say we have an image parameterization like before, with a shape of (5, 128, 128, 3) . Then we could ask for a lowres_tensor of shape (1, 128, 128, 3) to get a shared parameterization. Here, for instance, is the fixed parameterization area, each taking up one quarter of the image. It is now expanded to a pyramid of spaces in a manner reminiscent of a generalized Laplacian pyramid, as demonstrated in the code below. 20 Tool Computer Vision News Focus on… def lowres_param_f(): shared = lowres.lowres_tensor((6, 128, 128, 3), (1, 128//16, 128//16, 3)) return color.to_valid_rgb(shared, decorrelate=True)