Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Unsupervised Photometric-Consistent Depth Estimation from Endoscopic Monocular Video

Authors: Shijie Li, Weijun Lin, Qingyuan Xiang, Yunbin Tu, Shitan Asu, Zheng Li

AAAI 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate our method achieves the state-of-the-art results on C3VD, SCARED and SERV-CT datasets.
Researcher Affiliation Academia 1College of Computer Science, Sichuan University, Chengdu, China 2 School of Computer Science and Technology, University of Chinese Academy of Sciences Beijing, China EMAIL
Pseudocode No The paper describes methods using natural language and mathematical equations but does not include any explicitly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code https://github.com/DpEstimation/PC-Depth
Open Datasets Yes C3VD (Bobrow et al. 2023). The images are captured by using a genuine Olympus CF-HQ190L endoscope within a silicone model of a human colon. SCARED (Allan et al. 2021). The dataset is acquired using a da Vinci Xi endoscope on fresh porcine cadaver abdominal anatomy. SERV-CT (Edwards et al. 2022). SERV-CT includes 16 stereo pairs collected from ex vivo porcine torso cadavers, along with corresponding depth and disparity ground truth.
Dataset Splits Yes C3VD (Bobrow et al. 2023)... We allocate 8,690 frames for training, 148 for validation, and 2,888 are designated for testing purposes. SCARED (Allan et al. 2021)... the dataset is divided into a training set containing 15,351 frames, a validation set containing 1,705 frames, and a test set containing 551 frames.
Hardware Specification Yes Our PC-Depth is trained on one GTX 3090 GPU with a batch size of 4 for 20 epochs.
Software Dependencies No The paper mentions using 'adam optimizer' and 'Res Net-18' which are algorithms/architectures, but does not specify software dependencies like programming languages or libraries with version numbers (e.g., Python, PyTorch, TensorFlow versions).
Experiment Setup Yes Our PC-Depth is trained on one GTX 3090 GPU with a batch size of 4 for 20 epochs. Following (Bian et al. 2019), adam optimizer (Kingma and Ba 2014) is used with an initial learning rate of 1e-4 and drops to 1e-5 after 10 epochs. We utilize the pre-trained weights on Image Net (Deng et al. 2009) for Res Net initialization. We adopt α = 1.0, β = 0.01, γ = 0.1 and ω = 0.1 in Equation 1.