Table 4 Identification accuracy on manually and computed transformed WIPs pictures, modified pictures taken in the training dataset (A) and the test dataset (B)

From: Wing Interferential Patterns (WIPs) and machine learning, a step toward automatized tsetse (Glossina spp.) identification

  1. Legends: Green 100% match of identification, yellow identification below 100%, red no match of identification, white no samples submitted for identification. Transformation process using computer application (1 no transformation, 2 gaussian blurs, 3 lens blur, 4 video distortion, 5 scatter, 6 throws, 7 scrambling RVB, 8 scrambling CIE) or manual transformation (9 trailing edge cropping, 10 wings cropping from vein V, 11 wings cropping to the vein VI, 12 original picture without color saturation, contrast and resizing transformations, 13 horizontal flips, 14 vertical flips, 15 resized original picture).
Back to article page