CycleMLP: A MLP-like Architecture for Dense Prediction
Authors: Shoufa Chen, Enze Xie, Chongjian GE, Runjian Chen, Ding Liang, Ping Luo
ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on various tasks (e.g., Image Net classification, COCO object instance detection, and segmentation, and ADE20K semantic segmentation) demonstrate that Cycle MLP outperforms existing MLP-like models and is comparable to and sometimes better than CNNs and Transformers on dense predictions. |
| Researcher Affiliation | Collaboration | 1 The University of Hong Kong 2 Sense Time Research 3 Shanghai AI Laboratory, Shanghai, China {shoufach, xieenze, rhettgee, rjchen}@connect.hku.hk liangding@sensetime.com pluo@cs.hku.hk |
| Pseudocode | No | The paper describes the mathematical formulation of Cycle FC and the overall architecture in text, but it does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is available at https: //github.com/Shoufa Chen/Cycle MLP. |
| Open Datasets | Yes | In this section, we first examine Cycle MLP by conducting experiments on Image Net-1K (Deng et al., 2009) image classification. Then, we present a bunch of baseline models achieved by Cycle MLP in dense prediction tasks, i.e., COCO (Lin et al., 2014) object detection, instance segmentation, and ADE20K (Zhou et al., 2017) semantic segmentation. |
| Dataset Splits | Yes | We train our models on the Image Net-1K dataset (Deng et al., 2009), which contains 1.2M training images and 50K validation images evenly spreading 1,000 categories. (...) We conduct object detection and instance segmentation experiments on COCO (Lin et al., 2014) dataset, which contains 118K and 5K images for train and validation splits. (...) We conduct semantic segmentation experiments on ADE20K (Zhou et al., 2017) dataset, which covers a broad range of 150 semantic categories. ADE20K contains 20K training, 2K validation and 3K testing images. |
| Hardware Specification | Yes | All models are trained for 300 epochs on 8 Tesla V100 GPUs with a total batch size of 1024. (...) All models are trained on 8 Tesla V100 GPUs with a total batch size of 16 for 12 epochs (...) trains models for 40K iterations on 8 Tesla V100 GPUs with 4 samples per GPU. |
| Software Dependencies | No | The paper mentions software like Py Torch, timm, mmdetection, and mmsegmenation, but it does not specify their version numbers. |
| Experiment Setup | Yes | The experimental settings for Image Net classification are mostly from Dei T (Touvron et al., 2020), Swin (Liu et al., 2021b). The detailed experimental settings for Image Net classification can be found in Appendix E.1. (...) The optimizer is Adam W (Loshchilov & Hutter, 2017) with the momentum of 0.9 and weight decay of 5 10 2 by default. The cosine learning rate schedule is adopted with the initial value of 1 10 3. All models are trained for 300 epochs on 8 Tesla V100 GPUs with a total batch size of 1024. |