CNN: How to implement features obtained by global max pooling into a regression later as output images?

3 visualizaciones (últimos 30 días)
Sungkun
Sungkun el 1 de Jul. de 2020
Editada: Madhav Thakker el 22 de Dic. de 2020
Hi,
I have modified the CNN structure, using the regression layer. As usual, we have to use the same size of images for input and output. As shown in the attachment, the actual image size should be 32x32 with 3 channels. But I want to use features obtained by global maximum pooling (GMP) for the input and output. All features would be 1x1x128.
Is there any way to implement 1x1x128 features acquired by GMP into the regression layer as output images?

Iniciar sesión para responder a esta pregunta.

Respuestas (0)

Categorías

Más información sobre Object Detection en Help Center y File Exchange.

Etiquetas

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by