Unsupervised Feature Learning via Non-Parametric Instance Discrimination

Neural net classifiers trained on data with annotated class labels can also capture apparent visual similarity among categories without being directed to do so. We study whether this observation can be extended beyond the conventional domain of supervised learning: Can we learn a good feature representation that captures apparent similar- ity among instances, instead of classes, by merely asking the feature to be discriminative of individual instances? We formulate this intuition as a non-parametric classification problem at the instance-level, and use noise-contrastive estimation to tackle the computational challenges imposed by the large number of instance classes. Our experimental results demonstrate that, under unsupervised learning settings, our method surpasses the state-of-the-art on ImageNet classification by a large margin. Our method is also remarkable for consistently improving test performance with more training data and better network architectures. By fine-tuning the learned feature, we further obtain competitive results for semi-supervised learning and object detection tasks. Our non-parametric model is highly compact: With 128 features per image, our method requires only 600MB storage for a million images, enabling fast nearest neighbour retrieval at the run time.

This page is based on documents in MMSelfSup.

The classification benchmarks includes 4 downstream task datasets, VOC, ImageNet, iNaturalist2018 and Places205. If not specified, the results are Top-1 (%).

The Best Layer indicates that the best results are obtained from which layers feature map. For example, if the Best Layer is feature3, its best result is obtained from the second stage of ResNet (1 for stem layer, 2-5 for 4 stage layers).

Besides, k=1 to 96 indicates the hyper-parameter of Low-shot SVM.

The Feature1 - Feature5 don't have the GlobalAveragePooling, the feature map is pooled to the specific dimensions and then follows a Linear layer to do the classification. Please refer to r50_mhead_sz224_4xb64_step_ep90.py for details of config.

The AvgPool result is obtained from Linear Evaluation with GlobalAveragePooling. Please refer to r50_linear_sz224_4xb64_step_ep100.py for details of config.

The Feature1 - Feature5 don't have the GlobalAveragePooling, the feature map is pooled to the specific dimensions and then follows a Linear layer to do the classification. Please refer to r50_mhead_sz224_4xb64_step_ep28 for details of config.

The detection benchmarks includes 2 downstream task datasets, Pascal VOC 2007 + 2012 and COCO2017. This benchmark follows the evluation protocols set up by MoCo.

Please refer to faster_rcnn_r50_c4_mstrain_24k_voc0712.py for details of config.

Please refer to mask_rcnn_r50_fpn_mstrain_1x_coco.py for details of config.

The segmentation benchmarks includes 2 downstream task datasets, Cityscapes and Pascal VOC 2012 + Aug. It follows the evluation protocols set up by MMSegmentation.

Please refer to fcn_r50-d8_512x512_20k_voc12aug.py for details of config.

@inproceedings{wu2018unsupervised,
  title={Unsupervised feature learning via non-parametric instance discrimination},
  author={Wu, Zhirong and Xiong, Yuanjun and Yu, Stella X and Lin, Dahua},
  booktitle={CVPR},
  year={2018}
}