This is Cloud and ML Project with Jyothir S V & Zuhaib Akhtar.
In this work, we try to optimize VOLO by several methods to accelerate the training and make it perform better.
We modified VOLO architecture, achieve 95.72 with only 27M parameters. Here is the model file: https://drive.google.com/file/d/1CimTxXOBBkstpGfoiQuiFmxigzugJ9AS/view?usp=sharing