Implementation of "FishNet: A Versatile Backbone for Image, Region, and Pixel Level Prediction" [https://arxiv.org/abs/1901.03495]
Task: Image Classification
Dataset: CIFAR-10 [https://www.cs.toronto.edu/~kriz/cifar.html]
to train, do python main.py
--root_dir: root directory path where train, val, and test data folder exist
main.py line 97-99: path where dataloader is saved
train.py line 120, 123: path where best validation model is saved
best test achieved currently using fishnet150
batch_size=64, optimizer AdamW lr 1e-4, augmentation random grayscale and random horizontal flip
no pooling before first layer of fish tail (input res 32x32)
Accuracy: 87.5%
F1 weighted: 87.5%
F1 per class:
airplane 88.7%
automobile 93.5%
bird 82.4%
cat 75.6%
deer 85.7%
dog 81.6%
frog 90.6%
horse 91.3%
ship 93.3%
truck 91.7%