Image Recognition & Assessment

The Ask: "Can you automate our assessment process?"

The Background:
This partner used periodic visual inspections of their crops to determine the performance of their current batch. This was a manual assessment that was based on photographs. Individuals would use the photographs to identify the growth stage and size of the crops. A large number of images were taken to ensure the results of the assessment were representative. This assessment required significant manual effort and was introduced inherent user bias (due to some subjective parts of the assessment). They approached us to determine if we could automate their process.

The Output:
We assessed their current process through analysis of past data and speaking with their growers. We developed a software application that automatically processed images that were submitted. The application identified the crop from the background and assessed the relevant features; collating the assessment results in a database.
An Artificial Neural Network (ANN) was developed to perform some of the more subjective aspects of the assessment. The ANN was initially trained on the back data (photographs and assessments) the partner had available. A dashboard was also developed to facilitate manual user assessment to further train and refine the ANN going forwards.

The Results:
The partner's manual assessment time was drastically reduced <5% of original. Allowing the growers to spend more time growing and less time assessing.