Machine learning techniques are used to explore the performance of boosted top quark tagging treating jets as images. Tagging performances are studied in both hadronic and leptonic channels employing convolutional neutral networks (CNN) based analysis along with boosted decision trees (BDT). It is observed that overall tagging performance in leptonic channel is better than the case of hadronic. This computer vision approach is also applied to distinguish the left and right polarized top quarks, and a measurable event asymmetry is constructed to estimate the polarization. Results indicates that the CNN based classifier is more sensitive than the standard kinematic variables to top quark polarization. It is observed that, overall tagging performance in leptonic channel is better than the case of hadronic and also serves as a better probe to study polarization.