This paper presents a method for offline training of neural networks to track novel objects at 100 fps. The proposed tracker, called GOTURN, uses a simple feed-forward network with no online training required. It learns a generic relationship between object motion and appearance, allowing it to track novel objects without fine-tuning. The tracker is significantly faster than previous methods that use neural networks for tracking, which are typically very slow to run and not practical for real-time applications. The tracker is trained entirely offline using a combination of videos and images, and it can track objects at real-time speeds. The performance of the tracker improves as more videos are added to the training set. The tracker is the first neural-network tracker that learns to track generic objects at 100 fps. The paper also compares the performance of the tracker with other state-of-the-art trackers on a standard tracking benchmark. The results show that the tracker outperforms previous methods in terms of accuracy and robustness. The tracker is able to track objects in real-time due to two factors: first, it is trained offline, so no online training is required; second, it uses a regression-based approach, requiring just a single feed-forward pass through the network to regress directly to the location of the target object. The combination of offline training and one-pass regression leads to a significant speed-up compared to previous approaches and allows the tracker to run at real-time speeds. The paper also discusses the advantages of offline training over online training, including faster training and better generalization to novel objects. The tracker is able to track objects in real-time due to its offline training and regression-based approach. The paper also discusses the results of experiments on a standard tracking benchmark, showing that the tracker outperforms previous methods in terms of accuracy and robustness. The results indicate that the tracker is able to track objects in real-time and is the first neural-network tracker to achieve this. The paper also discusses the limitations of the tracker, including its inability to handle fast-moving objects and occlusions. The paper concludes that the proposed method is effective for tracking novel objects at real-time speeds and that offline training is a promising approach for improving tracking performance.This paper presents a method for offline training of neural networks to track novel objects at 100 fps. The proposed tracker, called GOTURN, uses a simple feed-forward network with no online training required. It learns a generic relationship between object motion and appearance, allowing it to track novel objects without fine-tuning. The tracker is significantly faster than previous methods that use neural networks for tracking, which are typically very slow to run and not practical for real-time applications. The tracker is trained entirely offline using a combination of videos and images, and it can track objects at real-time speeds. The performance of the tracker improves as more videos are added to the training set. The tracker is the first neural-network tracker that learns to track generic objects at 100 fps. The paper also compares the performance of the tracker with other state-of-the-art trackers on a standard tracking benchmark. The results show that the tracker outperforms previous methods in terms of accuracy and robustness. The tracker is able to track objects in real-time due to two factors: first, it is trained offline, so no online training is required; second, it uses a regression-based approach, requiring just a single feed-forward pass through the network to regress directly to the location of the target object. The combination of offline training and one-pass regression leads to a significant speed-up compared to previous approaches and allows the tracker to run at real-time speeds. The paper also discusses the advantages of offline training over online training, including faster training and better generalization to novel objects. The tracker is able to track objects in real-time due to its offline training and regression-based approach. The paper also discusses the results of experiments on a standard tracking benchmark, showing that the tracker outperforms previous methods in terms of accuracy and robustness. The results indicate that the tracker is able to track objects in real-time and is the first neural-network tracker to achieve this. The paper also discusses the limitations of the tracker, including its inability to handle fast-moving objects and occlusions. The paper concludes that the proposed method is effective for tracking novel objects at real-time speeds and that offline training is a promising approach for improving tracking performance.