Instance Segmentation and Fruit Recognition In Colored Images using CNNs
This video proposes a way to recognize several kinds of fruits in an image. This project is divided into 3 parts - Fruit Recognition using Convolutional Neural Networks (CNN), Instance Segmentation and extraction of sub-images of fruits from large images containing multiple fruits and finally predicting the fruit labels of the sub-images extracted from the previous part using the model trained earlier on Fruits-360 dataset. An approach to identify different fruits in Kaggle's Fruits-360 dataset using CNN has been discussed. Alternative methods of Instance Segmentation with limitations are discussed in further sections. Finally, predicting different fruits and challenges associated with the same are discussed in detail along with the results. Future scope and improvements have been discussed to further improve the performance and usability in real life applications.