Specific operations
Hardware supports for Deep Neural Networks
Article REF: H1098 V1
Specific operations
Hardware supports for Deep Neural Networks

Author : Daniel ETIEMBLE

Publication date: August 10, 2021 | Lire en français

Logo Techniques de l'Ingenieur You do not have access to this resource.
Request your free trial access! Free trial

Already subscribed?

5. Specific operations

5.1 Convolutions

The convolution principle is illustrated in figure 13 . A 2*2 filter is used with an activation map of size 3*3. Starting from the top left corner of the activation map, the filter is applied to the activation map and generates a value placed at the top corner of the result map. The filter is then shifted one position to the right for the next value on the result map. When the filter reaches the last placement, we start again from the next line. With a 2*2 filter on a 3*3 activation card, we obtain a 2*2 result activation card. This example uses a filter offset step of 1. A step greater than 1 can be used.

You do not have access to this resource.
Logo Techniques de l'Ingenieur

Exclusive to subscribers. 97% yet to be discovered!

You do not have access to this resource. Click here to request your free trial access!

Already subscribed?


Article included in this offer

"Software technologies and System architectures"

( 227 articles )

Complete knowledge base

Updated and enriched with articles validated by our scientific committees

Services

A set of exclusive tools to complement the resources

View offer details