1. Baseline(Performance: Oxford5k)
1.1 VGG16(TensorFlow Pre-trained Model)
POOL5: 58.21 - FC6: 61.03 - FC7: 55.54
1.2 VGG16 fine tuned on dataset CNN Image Retrieval Learns from BoW: Unsupervised Fine-Tuning with Hard Examples
POOL5: 0.679571 - FC6: 0.643479 - FC7: 0.569642