This is part of Udacity's Pytorch Facebook Challenge where we are required to be able to classify the flower breeds into 102 classes with highest accuracy possible

• Using Pretrained Resnet152 with all layers but the last layer(linear classifier) frozen (Validation accuracy:~93%)
• Using Pretrained Resnet152 with the last 3 layers unfrozen (Validation accuracy:~98.8%)
• Using Pretrained Resnet152 with the last 4 layers unfrozen (Validation accuracy:~99.1%)
• Same config as previous but adding using Adam with weight decay (Validation accuracy: ~99.5%) (This is the procedure you will see in the file)

MIT