Dynamic Resolution Network
Authors: Mingjian Zhu, Kai Han, Enhua Wu, Qiulin Zhang, Ying Nie, Zhenzhong Lan, Yunhe Wang
NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We then conduct extensive experiments on several benchmark networks and datasets. |
| Researcher Affiliation | Collaboration | 1Zhejiang University. 2Huawei Noah s Ark Lab. 3State Key Lab of Computer Science, ISCAS & University of Chinese Academy of Sciences. 4School of Engineering, Westlake University. 5Institute of Advanced Technology, Westlake Institute for Advanced Study. 6University of Macau. 7BUPT. |
| Pseudocode | No | The paper describes methods in text and mathematical equations but does not include a structured pseudocode or algorithm block. |
| Open Source Code | Yes | Code will be available at https://gitee.com/mindspore/models/tree/master/research/cv/DRNet. |
| Open Datasets | Yes | Image Net-1K dataset (Image Net ILSVRC2012) [2] is a widely-used benchmark to evaluate the classification performance of neural networks, which consists of 1.28M training images and 50K validation images in 1K categories. ... For the license of Image Net dataset, please refer to http://www.image-net.org/download. |
| Dataset Splits | Yes | Image Net-1K dataset (Image Net ILSVRC2012) [2] is a widely-used benchmark to evaluate the classification performance of neural networks, which consists of 1.28M training images and 50K validation images in 1K categories. For data processing during validation, we first resize the input image into 256 256 and then crop the center 224 224 part. |
| Hardware Specification | Yes | The framework is implemented in Pytorch [21] on NVIDIA Tesla V100 GPUs. |
| Software Dependencies | No | The framework is implemented in Pytorch [21] on NVIDIA Tesla V100 GPUs. |
| Experiment Setup | Yes | For data augmentation during training for both Image Net-100 and Image Net-1K, we follow the scheme as in [6] including randomly cropping a patch from the input image and resizing to candidate resolutions with the bilinear interpolation followed by random horizontal flipping with probability 0.5. ... Optimization is performed using SGD (mini-batch stochastic gradient descent) and learning rate warmup is applied for the first 3 epochs. In the pretraining stage, the model is trained with total epochs 70, batch-size 256, weight decay 0.0001, momentum 0.9, initial learning rate 0.1 which decays a factor of 10 every 20 epochs. We adopt a similar training scheme in finetuning stage. The total epochs are 100 with the learning rate decaying a factor of 10 every 30 epochs. We adopt 1 learning rate to finetune the large classifier and 0.1 learning rate to train the resolution predictor from scratch. |