Source: Deep Learning on Medium
Seeing The World: Fruit and Vegetable Image Classifier
We started the Seeing the World project last year and landed the Microsoft Accessibility Grant to continue to build the project. As of today, we have collected 1833 fruit/vegetable pictures taken at Farmers’ Market/grocery stores. It covers 60 different categories of fruit and vegetable.
Since then, we had built an initial version of deep learning model to recognize these different categories of fruit and vegetable. The result looks promising and we continue to improve the model performance. The below shows the classification report of precision, recall, F1-score of the classifier. There are categories that need improvement due to insufficient number of images for training those categories.
Here are some sample images: