EAGLE: Efficient Adaptive Geometry-based Learning in Cross-view Understanding
Authors: Thanh-Dat Truong, Utsav Prabhu, Dongyi Wang, Bhiksha Raj, Susan Gauch, Jeyamkondan Subbiah, Khoa Luu
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experiments on different cross-view adaptation benchmarks have shown the effectiveness of our approach in cross-view modeling, demonstrating that we achieve State-of-the-Art (SOTA) performance compared to prior unsupervised domain adaptation and open-vocabulary semantic segmentation methods. |
| Researcher Affiliation | Collaboration | Thanh-Dat Truong1, Utsav Prabhu2, Dongyi Wang3 Bhiksha Raj4,5, Susan Gauch6, Jeyamkondan Subbiah7, Khoa Luu1 1CVIU Lab, University of Arkansas, USA 2Google Deep Mind, USA 3Dep. of BAEG, University of Arkansas, USA 4Carnegie Mellon University, USA 5Mohammed bin Zayed University of AI, UAE 6Dep. of EECS, University of Arkansas, USA 7Dep. of FDSC, University of Arkansas, USA {tt032, dongyiw, sgauch, jsubbiah, khoaluu}@uark.edu bhiksha@cs.cmu.edu, utsavprabhu@google.com |
| Pseudocode | No | The paper describes its methodology using mathematical formulations and textual descriptions but does not include any explicit pseudocode or algorithm blocks. |
| Open Source Code | No | The code will be published may the paper be accepted. |
| Open Datasets | Yes | Following common practices in UDA [23, 58], we choose SYNTHIA [44], GTA [43], and BDD100K [68] as the source domains while UAVID [33] is chosen as the target domain. |
| Dataset Splits | No | The paper mentions using source and target domains for training and evaluation but does not provide explicit details about training, validation, and test dataset splits with percentages or sample counts. |
| Hardware Specification | No | The paper generally mentions the use of 'GPUs' and 'GPU resources' for experiments but does not provide specific details such as GPU models (e.g., NVIDIA A100), CPU types, or memory specifications. |
| Software Dependencies | No | The paper mentions software components such as Mask2Former, Free Seg, ResNet, Swin, and CLIP, but it does not provide specific version numbers for these software dependencies, which are necessary for full reproducibility. |
| Experiment Setup | Yes | The linear scale factors α and γ are set to α = 1.5 and γ = 1.0, respectively. For the Geodesic Flow modeling, we adopt the implementation of generalized SVD decomposition [18, 47] in the framework. The subspace dimension in our geodesic flow-based metrics is set to D = 256. The batch size and the base learning rate in our experiments are set to 16 and 2.5e-4. The balanced weights of losses in our experiments are set to λI = 1.0 and λP = 0.5. |