Group Member: Mu Cai, Yunyu(Bella) Bai, Xuechun Yang
Source code and dataset available at: https://github.com/mu-cai/cs766_21spring
Related Work
The Traditional DL to Train a Model: ERM
![ERM.png](https://static.wixstatic.com/media/3c7a9b_42ed2f6e4eeb468998665759a34bbee6~mv2.png/v1/fill/w_940,h_376,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/ERM.png)
State-of-the-art Methods that Reduce Group Shifts: DRO
![DRO 2.png](https://static.wixstatic.com/media/3c7a9b_825aaf4b495a4b55bfdfc30b80d5b408~mv2.png/v1/fill/w_45,h_22,al_c,q_85,usm_0.66_1.00_0.01,blur_2,enc_auto/DRO%202.png)
The Performance of the Above Two Methods
Even though DRO focus on minimizing the worst-group loss, the test accuracy of the minor groups is still far from that of the major group!
![Screen Shot 2021-04-29 at 11.35.10 PM.pn](https://static.wixstatic.com/media/3c7a9b_e0c0b4a3cb9c41cb87f658a5f276a2e1~mv2.png/v1/fill/w_47,h_10,al_c,q_85,usm_0.66_1.00_0.01,blur_2,enc_auto/Screen%20Shot%202021-04-29%20at%2011_35_10%20PM_pn.png)
Current Limitation 1: Synthesized Dataset
The current community utilizes fake data to construct the group-shifted dataset, which doesn’t reflect the distribution of natural images, blocking its real-world applications.
As shown in the picture below, this dataset is constructed by simply stitching a background image and a foreground object.
To facilitate the research in group shifts for the community, we collect a large-scale natural image dataset via web-crawler.
![Limitation 1.png](https://static.wixstatic.com/media/3c7a9b_bf29282c0e0f448ab30fd25f6aa6c8ca~mv2.png/v1/fill/w_47,h_10,al_c,q_85,usm_0.66_1.00_0.01,blur_2,enc_auto/Limitation%201.png)
Current Limitation 2: OOD Dataset
Besides, the current research community also doesn’t consider its robustness towards out-of-distribution samples. Real world test images has a wide span of distribution. Therefore, determining whether test images belong to the in-distribution set is critical, which is not yet studied in the community. Here we study the robustness of the neural network models under four diverse high resolution out-of-distribution datasets.
![Limitation 2.png](https://static.wixstatic.com/media/3c7a9b_1ef776a5aad5472b81852565a6bdc2bb~mv2.png/v1/fill/w_47,h_13,al_c,q_85,usm_0.66_1.00_0.01,blur_2,enc_auto/Limitation%202.png)