GitHub - arf-themascoteers/glasses

#Base Dataset

#Setup

Download the dataset and put all the jpg images in the data/faces folder

pip install -r requirements.txt

#Run

Run main.py

#Transfer Learning

Facts:

Keep the penultimate layer
Freeze convolutional blocks
Depth augmentation (two new layers beyond pre-trained classification layer is cut-off point)
Layer-wise fine tuning (not implemented)
Normalisation/Regularisation - normalize and scale input activations of augmented layers

##Referred Papers

Oquab, M., Bottou, L., Laptev, I., Sivic, J.: Learning and transferring mid-level image representations using convolutional neural networks. Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1717{1724 (2014). (Discarding penultimate)
Wang, Y., Ramanan, D., Hebert, M.: Growing a brain: Fine-tuning by increasing model capacity. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2471{2480 (2017). (Discarding penultimate, L2-norm)
S., Szegedy, C.: Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift. Proceedings of the 32Nd International Conference on International Conference on Machine Learning, vol-37 (Batch-norm)

Name		Name	Last commit message	Last commit date
Latest commit History 26 Commits
data		data
models		models
.gitignore		.gitignore
Pipfile		Pipfile
README.md		README.md
cnn.py		cnn.py
face_read.py		face_read.py
glasses_dataset.py		glasses_dataset.py
machine_provider.py		machine_provider.py
main.py		main.py
our_machine.py		our_machine.py
playground.py		playground.py
requirements.txt		requirements.txt
saliency_map.py		saliency_map.py
test.py		test.py
traditional_machine.py		traditional_machine.py
train.py		train.py
trans.png		trans.png

arf-themascoteers/glasses