In this project, I used Python + a caffe-trained CNN to classify and localize images from a subset of ImageNet. This project is divided into two parts:
-
Image Classification
In this part, a CNN which takes a 32x32x3 image as input and has 4 output nodes followed by a Softmax layer is designed and trained.
-
Image Localization
In this part, some pre and post processing is done in order to localize a class instance in an image. This is implemented in four phases:
Phase 1: Creation of dataset for training Phase 2: Preparing train and test splits Phase 3: Training Phase 4: Testing
The NN created in Part1 is used in Part2 as well.
This project is done as a part of "Artificial Intelligence" course, fall 2017, Purdue University