A new dataset, DRENDS, addresses depth estimation in robotic minimally invasive surgery (RMIS), where analyzing laparoscopic images to obtain reliable depth can enable force estimation, 3D reconstruction, augmented reality, and autonomy. The integration of active sensors like LiDAR and Time-of-Flight sensors is often costly and complex due to the limited workspace in RMIS procedures, which demands small, low-range sensors. Passive RGB cameras offer a solution but rely on data with 3D ground truth for algorithm validation and training. Existing public datasets have limitations: SERV-CT comprises 16 stereo-endoscopic image pairs from ex vivo porcine samples; the Hamlyn Laparoscopic/Endoscopic Dataset provides video sequences with ground-truth depth maps generated using Libelas stereo matching software; EndoAbs offers 120 non-sequential, low-resolution frames (640×480) of 3D-printed abdominal organ phantoms under challenging conditions; and SCARED provides nine samples, each with 4 to 5 short stereo frame sequences, where the first frame has high-precision depth from a structured light approach.