Semi-Supervised Recognition under a Noisy and Fine-grained Dataset
Simi-Supervised Recognition Challenge-FGVC7 is a challenging fine-grained recognition competition. One of the difficulties of this competition is how to use unlabeled data. We adopted pseudo-tag data mining to increase the amount of training data. The other one is how to identify similar birds with a very small difference, especially those have a relatively tiny main-body in examples. We combined generic image recognition and fine-grained image recognition method to solve the problem. All generic image recognition models were training using PaddleClas . Using the combination of two different ways of deep recognition models, we finally won the third place in the competition.
PDF AbstractCode
Datasets
ImageNet
Results from the Paper
Ranked #242 on
Image Classification
on ImageNet
(using extra training data)
| Task | Dataset | Model | Metric Name | Metric Value | Global Rank | Uses Extra Training Data |
Benchmark |
|---|---|---|---|---|---|---|---|
| Image Classification | ImageNet | MobileNetV3_large_x1_0_ssld | Top 1 Accuracy | 79.0% | # 727 | ||
| Number of params | 5.47M | # 421 | |||||
| Image Classification | ImageNet | ResNet200_vd_26w_4s_ssld | Top 1 Accuracy | 85.1% | # 245 | ||
| Number of params | 76M | # 799 | |||||
| Hardware Burden | None | # 1 | |||||
| Operations per network pass | None | # 1 | |||||
| Image Classification | ImageNet | Fix_ResNet50_vd_ssld | Top 1 Accuracy | 84.0% | # 336 | ||
| Number of params | 25.58M | # 599 | |||||
| Image Classification | ImageNet | ResNet50_vd_ssld | Top 1 Accuracy | 83.0% | # 437 | ||
| Number of params | 25.58M | # 599 |
