This paper addresses the problem of tracking an object in a video given only its initial location, without any prior information. The authors propose using Multiple Instance Learning (MIL) instead of traditional supervised learning to train a discriminative classifier for object tracking. This approach avoids issues with incorrect labeling due to slight inaccuracies in the tracker, leading to more robust and stable tracking. The paper introduces a novel online MIL algorithm, MILTrack, which updates the appearance model using multiple positive examples and negative examples from the current frame. Experimental results on challenging video clips demonstrate that MILTrack outperforms existing methods in terms of robustness and stability, especially under partial occlusions and appearance changes. The paper also discusses the implementation details and evaluates the performance of MILTrack compared to other state-of-the-art trackers.This paper addresses the problem of tracking an object in a video given only its initial location, without any prior information. The authors propose using Multiple Instance Learning (MIL) instead of traditional supervised learning to train a discriminative classifier for object tracking. This approach avoids issues with incorrect labeling due to slight inaccuracies in the tracker, leading to more robust and stable tracking. The paper introduces a novel online MIL algorithm, MILTrack, which updates the appearance model using multiple positive examples and negative examples from the current frame. Experimental results on challenging video clips demonstrate that MILTrack outperforms existing methods in terms of robustness and stability, especially under partial occlusions and appearance changes. The paper also discusses the implementation details and evaluates the performance of MILTrack compared to other state-of-the-art trackers.