Skip to main content
. 2018 Mar 23;14:25. doi: 10.1186/s13007-018-0292-9

Fig. 2.

Fig. 2

A schematic of the CNN architecture employed for wood identification. We trained models with both global average pooling and global max pooling layers (with the performance being comparable). The dimensions of the feature maps are in pixels of the form: (height, width, depth). The final classification layers has 10 and 6 outputs for the species and genus level models respectively