Abstract--Surgical suturing is a high-precision task that impacts patient healing and scarring. Suturing skill varies widely between surgeons, highlighting the need for robot assistance. Previous robot suturing works, such as STITCH 1.0 [1], struggle to fully close wounds due to inaccurate needle tracking and poor thread management. T o address these challenges, we present STITCH 2.0, an elevated augmented dexterity pipeline with seven improvements including: improved EKF needle pose estimation, new thread untangling methods, and an automated 3D suture alignment algorithm. Experimental results over 15 trials find that STITCH 2.0 on average achieves 74.4% wound closure with 4.87 sutures per trial, representing 66% more sutures in 38% less time compared to the previous baseline. When two human interventions are allowed, STITCH 2.0 averages six sutures with 100% wound closure rate. URGICAL robots have revolutionized minimally invasive surgery, with Intuitive Surgical's da Vinci system performing over 2.6 million procedures in 2024 [2]. While these procedures require complete human control, recent advances in artificial intelligence (AI) present opportunities for surgical robot autonomy. However, the high-risk nature of surgery raises safety concerns for fully autonomous AI systems.

Abstract--Deep perceptual loss is a type of loss function in computer vision that aims to mimic human perception by using the deep features extracted from neural networks. In recent years, the method has been applied to great effect on a host of interesting computer vision tasks, especially for tasks with image or image-like outputs, such as image synthesis, segmentation, depth prediction, and more. Despite the increased interest and broader use, more effort is needed toward exploring which networks to use for calculating deep perceptual loss and from which layers to extract the features. This work aims to rectify this by systematically evaluating a host of commonly used and readily available, pretrained networks for a number of different feature extraction points on four existing use cases of deep perceptual loss. The use cases of perceptual similarity, super-resolution, image segmentation, and dimensionality reduction, are evaluated through benchmarks. The procedure followed in this work. In the last decade, machine learning for computer vision One of the methods that makes use of loss networks that have has evolved significantly. This evolution is mainly due to proven effective for a range of computer vision applications the developments in artificial neural networks. An essential is deep perceptual loss. Deep perceptual loss aims to create focus of these developments has been the calculation of the loss functions for machine learning models that mimic human loss used to train models. One group of loss functions build perception of similarity. This is typically done by calculating on using a neural network to calculate the loss for another the similarity of the output image and ground truth by how machine learning model. Among these methods are milestone similar the deep features (activations) of the loss network are achievements such as adversarial examples [1], generative when either image is used as input. Benchmark 3 [7] Deep perceptual loss is well suited for image synthesis, where evaluates U-nets [21] trained with the loss networks on the it is utilized to improve performance on various tasks.