smartphone imaging analysis (1)

10910031863?profile=RESIZE_400x

Saffron is a high value spice and hence susceptible to adulteration and fraud. In this study, a machine vision system based on smartphone image analysis and deep learning was used to detect saffron authenticity and quality. A dataset of 1869 images was created of 6 types of saffron/adulterants including: dried saffron stigma using a dryer; dried saffron stigma using pressing method; pure stems of saffron; sunflower; saffron stems mixed with food colouring; and corn silk mixed with food colouring. The deep learning system developed for grading and authenticity determination of saffron in images captured by smartphones and applied to these images, was a Learning-to-Augment incorporated Inception-v4 Convolutional Neural Network (LAII-v4 CNN). After applying further data augmentation and comparison against regular CNN-based methods and traditional classifiers, the results showed that the proposed LAII-v4 CNN approach gave an accuracy of 99.5%.

 Read the paper preview here

Read more…