Skip to main content
Fig. 10 | Plant Methods

Fig. 10

From: TasselNetv2: in-field counting of wheat spikes with context-augmented local regression networks

Fig. 10

Some ground truth density maps overlaid on original images on the test set of the WSC dataset and count maps generated by TasselNetv2 (finetuned with pre-trained VGG16). The number above each original image denotes the ground truth count number of wheat spikes, while that above each density map denotes prediction count number. The last line shows some unsuccessful predictions, and error maps of these images are also presented. An error map denotes the difference of the ground truth and predicted density map. Over-estimate is denoted by red, under-estimate by blue, and minor difference by gray. The darker the color is, the greater the errors are. We also zoom in some local areas with high counting errors. ’GT’ denotes ground-truth counts and ’Error’ denotes the difference compared to the ground truth. Further visualizations can be found in Additional file 1.

Back to article page