Given a dataset of images, I need to segment foreground objects from the background for each image. the dataset is images of "Cars" . The output image should be just the car without any background from the original image. Can some please guide me to what are the broader steps needed to train this model ? What would be the algorithms needed . I will be using tensorflow as implementation . An example of desired functionality can be found at https://www.remove.bg/
Deep learning based semantic segmentation can be used to segment foreground from the background. You need annotated data like this http://cocodataset.org/#explore for training the DL model. VGG16 based semantic segmentation can extract precise edges of the foreground objects.