Fig. 9From: TasselNetv2: in-field counting of wheat spikes with context-augmented local regression networksThe distribution of absolute errors for local patches and test images. The left is the histogram of absolute error for local patches, and the right is the histogram of absolute error for test images. All networks are trained from scratch. “TasselNet (add-c)” denotes adding the context in TasselNet as per Fig. 6 since the training phaseBack to article page