The following has been achieved in this project:
- Multiple DNN, CNN, Transfer Learning Models have been trained using PyTorch and alayzed on CIFAR 10 dataset.
- A web application has been created using Flask where you can access the 4 models having the highest accuracy on the test set.
- All code here has been implemented by Chirag Bhuvaneshwara.
Note: The web application has been tested only in debug mode by running it as: python Flask_Web_App.py Flask_Web_App.py is found inside CIFAR10_Flask_Web_APP directory.
Accuracy achieved:
Models | Test Accuracy |
---|---|
Simple DNN | 42.52% |
Simple CNN | 81.46% |
VGG19_bn Net | 91.06% |
Alex Net | 90.05% |
GoogLeNet | 95.78% |
ResNet34 | 96.42% |
Inception v3 | * |
*=> I was unable to train because of insufficient VRAM. Probably needs 16GB of VRAM.
Screenshots from the web app:
Contact Me:
- Name: Chirag Bhuvaneshwara
- Email: [email protected]