T-MARS: Improving Visual Representations by Circumventing Text Feature Learning

Open in new window