Deep Learning Patterns and Practices by Andrew Ferlitsch

Author:Andrew Ferlitsch [Andrew Ferlitsch]
Language: eng
Format: epub
Publisher: Manning Publications
Published: 2021-09-21T00:00:00+00:00

Figure 8.6 MobileNet v2 macro-architecture

8.2.2 Stem

The stem component is similar to v1, except after the initial 3 × 3 convolutional layer, it is not followed by a depthwise convolutional block as in v1 (figure 8.7). As such, the coarse-level feature extraction would have less representational power than the dual stack of 3 × 3 in v1. The authors don’t state why the reduction in representational power did not impact the model, which outperformed v1 in accuracy.

Figure 8.7 MobileNet v2 stem group

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.