Releasing code examples for training AlexNet using Keras

[rahuldu...@gmail.com]

Dear community,

Apologies for cross posting.

I would like to share my experiences (read code examples) of training AlexNet using Keras, for three specific scenarios :-

1. Training from scratch

2. Fine-tuning the pre-trained AlexNet

3. Extracting features from a specific layer

This could help you establish the baseline performance using a standard CNN model (AlexNet) for your research.

Github link : here

Blog entry  : here

I would appreciate any suggestions or constructive feedback. If you find any bugs/errors, please post them under issues, on the github project webpage. All code is released under MIT license. Feel free to modify/edit and redistribute.

Two specific resources which were very helpful for my project are - this article by Francois Chollet and this project by the Heuritech team. Many thanks to them!

Thanks,

Rahul Duggal

[keyin...@gmail.com]

Hi, Rahul,

Thank you so much for sharing your code and experience with us. I'm lucky to come across your post when I'm starting my new project.

I applied your code to my small dataset and it works pretty well. It will be great if you could update your code to keras 2 API.

I have a question. When we train the Alexnet from scratch, do your code initialize the weights? I couldn't find this information. Thank you, again.

Best,
Jane

[geetanj...@nitttrchd.ac.in]

not working this code on miniconda