This folder contains 3 subfolders to provide more feature visualizations for the research paper : An analysis of the transfer learning of convolutional neural networks for artistic images ; Gonthier N., Gousseau Y., Ladjal S. ; ICPR 2020 FAPER Workshop https://arxiv.org/abs/2011.02727
The first folder 1_Pairs_RASTA_finetuned_A1_pretrained contains pairs of feature visualization for all the channel of a certain number of layers of the InceptionV1 model. 
For each pair of images, the optimized image on the left is obtained with the model fine-tuned on the artistic style dataset RASTA. The image on the right is obtained with the pretrained ImageNet model.
2_ImageNet_pretrained contains the optimized images synthesized with the pretrained ImageNet model.
3_RASTA_finetuned_A1 contains the optimized images synthesized with the model fine-tuned on the artistic style dataset RASTA (Mode A training 1).

Those feature visualizations have been made thanks to the Lucid Framework : https://github.com/tensorflow/lucid