Modern quadruped robots commonly rely on visual feedback and proprioceptive sensors for control and environmental analysis. However, these sensors lack the ability to perceive fine level contact information with the environment. To overcome these challenges, this paper introduces an innovative solution that integrates vision-based soft tactile sensors with a quadruped robot, providing the robot with enriched environmental feedback while maintaining compliant interactions. We present the design and modelling of our proposed solution on a commercially available quadruped robot. Through a learning based modelling pipeline, we equip the robot to estimate the magnitude and direction of contact forces in interaction with the ground. We believe such tactile information can significantly enhance the robot’s perception and interaction capabilities, offering deeper insights into surface properties and improving its adaptability in diverse conditions.