Fast Neural Architecture Search for Lightweight Dense Prediction Networks
This work addresses the need for fast and compact neural networks for dense prediction tasks like depth estimation and segmentation, offering a practical solution for resource-constrained applications.
The authors tackled the problem of efficient neural architecture search for lightweight dense prediction networks, achieving consistent improvements across multiple tasks while reducing model size by 5% to 315% compared to prior methods.
We present LDP, a lightweight dense prediction neural architecture search (NAS) framework. Starting from a pre-defined generic backbone, LDP applies the novel Assisted Tabu Search for efficient architecture exploration. LDP is fast and suitable for various dense estimation problems, unlike previous NAS methods that are either computational demanding or deployed only for a single subtask. The performance of LPD is evaluated on monocular depth estimation, semantic segmentation, and image super-resolution tasks on diverse datasets, including NYU-Depth-v2, KITTI, Cityscapes, COCO-stuff, DIV2K, Set5, Set14, BSD100, Urban100. Experiments show that the proposed framework yields consistent improvements on all tested dense prediction tasks, while being $5\%-315\%$ more compact in terms of the number of model parameters than prior arts.