Dear ML expert,
I am working on a challenging task of classifying wide-face vs narrow-face. I have a database with 800+ samples in each category.
The validation-accuracy I could achieve with inception-V3 is 68%, without data augmentation. With data augmentation, it is at 79%. However this validation-accuracy is not sufficient because using the generated model, some of the test images fails.
If you can help me with this challenge, please bid. I will share the database so that you can give a try. The couple of requirements are as follows.
1) model size needed to be less than 85MB
2) Only Tensorflow can be used
3) All test images need to pass (10 of them)
4) validation-accuracy needed to be above 90%
5) It is not possible to get any more samples