Hi Everybody
This is an interesting article.
Regards Conwyn
Hi Conwyn hope all is well!
I read the artical! I not surprised this simple aproach beats bigger models trained with large datasets. However I sm surprised by their solution, using a model based on a primates cortex model.
One of the strengths of the VOneBlock is its compatibility with current CNN architectures. “The VOneBlock was designed to have a plug-and-play functionality,” Marques says. “That means that it directly replaces the input layer of a standard CNN structure. A transition layer that follows the core of the VOneBlock ensures that its output can be made compatible with the rest of the CNN architecture.”
It would be great to apply this to the bear classifier 
Cheers mrfabulous1 
Hi Mr F.
I was re-visiting the Deep Learning GAN, Cycle-GAN and Style transfer video and in Cycle-GAN you have a Horse so the model say (Horse) 100% true. You then convert the Horse into a Zebra. Model(Zebra(Horse)) 25% true (yes four legs) and the Model(Horse(Zebra(Horse)) 0.1% true so lets try again and eventually you can convert a picture of a Horse into a Zebra (stripy horse). So I decided to revisit the Bears but use a celebrity dataset which contains people with a multi-category spreadsheet. The images are 2GB so I filtered them down to 112 similar images with one category. The results were very poor with RESNET34 and to add to my disillusionment why I try predict(my image) it scored quite highly and I can assure you I am no celebrity. Now multi-category needs threshold to define out of category. Consequently I think I will revisit Bears as multi-category and see if it can differentiate between me and a bear. The point is even your average bear I believe can differentiate between bears and humans so primates including humans front ending deep learning is interesting but perhaps the humans choosing the contents to create RESNET and performance of white male images suggest it might be a unfortunate idea.