Latency-aware Spatial-wise Dynamic Networks

Authors: Yizeng Han, Zhihang Yuan, Yifan Pu, Chenhao Xue, Shiji Song, Guangyu Sun, Gao Huang

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on image classification, object detection and instance segmentation demonstrate that the proposed framework significantly improves the practical inference efficiency of deep networks.
Researcher Affiliation Academia Yizeng Han1 Zhihang Yuan2 Yifan Pu1 Chenhao Xue2 Shiji Song1 Guangyu Sun2 Gao Huang1 1 Department of Automation, BNRist, Tsinghua University, Beijing, China 2 School of Electronics Engineering and Computer Science, Peking University, Beijing, China
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/Leap Lab THU/LASNet.
Open Datasets Yes The image classification experiments are conducted on the Image Net [4] dataset. ... We further evaluate our LASNet on the COCO [22] object detection task.
Dataset Splits No The paper mentions evaluating on the 'Image Net validation set' and conducting experiments on the 'Image Net [4] dataset' and 'COCO [22] object detection task', which implies standard splits for these public datasets are used. However, it does not explicitly provide specific percentages or sample counts for training/validation/test splits within the main text.
Hardware Specification Yes Various types of hardware platforms are tested, including a server GPU (Tesla V100), a desktop GPU (GTX1080) and edge devices (e.g., Nvidia Nano and Jetson TX2).
Software Dependencies No The paper mentions using 'torchvision pre-trained models' but does not specify software dependencies like PyTorch, CUDA, or other libraries with version numbers.
Experiment Setup Yes We fix α = 10, β = 0.5 and T = 4.0 for all dynamic models. More details are provided in Appendix B. ... finetune the whole network for 100 epochs ... finetuned on COCO with the standard setting for 12 epochs