The problem SCENER.io solves

Create a platform for training, labeling and deploying and
retraining image classification models.
Following are the expected features to be deployed:

Dataset : https://www.kaggle.com/puneet6060/intel-image-classification
Train a model(s) that classifies images based on the
dataset provided.

Create a UI that can run inference using the model trained
above on 128 unseen and unlabeled images uploaded at
the same time.
Once inference is completed, the UI should then be able to
visualize these images and their predictions including the
confidence score and provide other metrics as appropriate.
The UI should have the functionality to change the labels of
images that are wrong, add them to a database and run
training again.
Optionally, the UI should have an option to change the
parameters of training. Parameters could be learning rate,
number of filters, filter size etc.
The newly trained model should be available for use by the
UI to run another round of inference.
Extra credit will be given if the entire process is done on the
cloud.

Challenges we ran into

Hosting project on the cloud using docker was difficult beacuse we used flask for our backend.