# 3rd Position

I have used timm(to include much effective models).I did ensembling of dla102,gluon_resnext101_32x4d,dla102x which gave me private score of 0.451940 and public score 0.44787.I am sharing my colab notebook of all the 3 models.
[dla102 model](https://colab.research.google.com/drive/1FrFF-Mtd-AA8KsLYp7E-eGFffDpj7Vtb)
[dla 102x model](https://colab.research.google.com/drive/1b06RaGvGUPzaVfxhLwJYX0zEW5HD9OIH)
[gluon_resnext101_32x4d model](https://colab.research.google.com/drive/19hEhd_LRZQgV5tjZkeX1E90P2yTJhL_x)
[Ensembler model](https://colab.research.google.com/drive/1TcTM3DSWs4mjxjgLP-RI9ium8EfkukqT)
The attached submission (6).csv is collection of all the outputs from different trained models and a.csv is the final ensembled output.