AudioEar: Single-View Ear Reconstruction for Personalized Spatial Audio
Authors: Xiaoyang Huang, Yanjun Wang, Yang Liu, Bingbing Ni, Wenjun Zhang, Jinxian Liu, Teng Li
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments Ear Reconstruction Benchmark Evaluation We leverage the full Audio Ear3D dataset to evaluate the performance of ear reconstruction models. We compute the distance from each point in the ground-truth ear scan to the predicted ear mesh. ... Results As listed in Table 3 #6-9, our method surpasses the average ear baseline, HERA and DECA-coarse, with a final S2M distance of 1.28 mm. |
| Researcher Affiliation | Collaboration | Xiaoyang Huang1, Yanjun Wang1, Yang Liu2, Bingbing Ni1*, Wenjun Zhang1, Jinxian Liu1, Teng Li3 1Shanghai Jiao Tong University, Shanghai 200240, China, 2Focus Media, 3Anhui University |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found. |
| Open Source Code | Yes | Code and data are publicly available in https://github.com/seanywang0408/Audio Ear. |
| Open Datasets | Yes | Code and data are publicly available in https://github.com/seanywang0408/Audio Ear. ... To benchmark the ear reconstruction task, we introduce Audio Ear3D, a highquality 3D ear dataset consisting of 112 point cloud ear scans with RGB images. ... To self-supervisedly train a reconstruction model, we further collect a 2D ear dataset composed of 2,000 images, each one with manual annotation of occlusion and 55 landmarks, named Audio Ear2D. To our knowledge, both datasets have the largest scale and best quality of their kinds for public use. |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, and test sets. It mentions 'testset' but not the process of splitting. |
| Hardware Specification | No | The paper mentions 'Mantis Vision F6-SR 2' 3D scanner for data collection but does not provide specific hardware details (like GPU/CPU models, processor types, or memory) used for running the computational experiments or training the models. |
| Software Dependencies | No | The paper mentions various tools and models (e.g., Yolov4, WHENet, COMSOLTM, Detri2 program) but does not provide specific version numbers for these or any other software dependencies, libraries, or programming languages used for the experiments. |
| Experiment Setup | Yes | We use Adam (Kingma and Ba) with an initial learning rate of 0.45 for optimization. Each stage lasts for 166 iterations. The learning rate is multiplied by 0.1 when entering the next stage. |