Segmenting objects in images with transformers

Most recent transformer-based models show impressive performance on vision tasks, even better than Convolution Neural Networks (CNN).