In this project (Tiny ImageNet visual recognition challenge), there are 200 different classes. The training data has 500 images per class, with 50 validation images and 50 test images, with the validation and training images provided with labels and annotations.
The problem statement requires participants to predict labels, without needing to annotate the test images.
In this project, I used PyTorch Ignite to simply the deep learning implementation, and leveraged the power of EfficientNet to train an image classification model.