Training the Mask-RCNN model on a new class of objects comprising o…

[Shubhi3199]

Training the Mask-RCNN model on a new class of objects comprising of Tom & Jerry Cartoon characters

The approach utilizes Mask R-CNN for cartoon character detection, segmentaion and then utilizes transfer learning for training VGG16 CNN for emotion classification in cartoon images.

Important Updates - Refer Code here

• TomJerry.py - main file to train the Mask-RCNN model on custom dataset (images & json files) comprising of Tom & Jerry Characters.
• Inspect_TomJerry_Model.ipynb - For visualizing the cartoon character face masks generated by the Mask-RCNN model on Tom-Jerry Dataset.
• Tom_&_Jerry_TF_2.1_Combined.ipynb - Comprises of the complete flow of the project in a single ipynb file, which includes :
  • Detection of Cartoon character in the input frame
  • Segmentation of the cartoon charatcer's face mask
  • Training VGG16 model for classifying the emotion of the cartoon charatcer in the image into 4 categories namely, 'Happy', 'Sad', 'Angry', and 'Surprise'.