Assignment 1 (Pick one A or B):

A: Train a Computer Vision based Deep Learning Model and Develop a Web Application to Inference it.

Overview:

In this assignment, you will be developing a training module for image recognition which uses pre-trained deep learning model and then fine-tune it with the given dataset. Do required pre-processing on the given dataset. Also, you will be developing a web application that integrates the fine-tuned model and recognize the image uploaded by the user.

Requirements:

Training Module:

* Pre-process the given dataset.

* Fine tune the pre-trained model with given dataset.   

Users should be able to do the following in the Web App:

* Upload an image.

* Pre-process the image for the deep learning model.

* Predict the class of the uploaded image using the fine tuned model.

* View the predicted class label and probability.

Justify the selection of pre-trained model, model architecture and Pre-process on the provided dataset to the Hackthon panel.

Good to Have

Dockerize the solution and deploy using docker-compose.

Submission:

• Submit the source code for the web application.
• Submit a brief report explaining your model selection justification.

B: Train a Natural Language Processing based Deep Learning Model and Develop a Web Application to Inference it.

Overview:

In this assignment, you will be developing a training module for text classification which uses pre-trained deep learning model and then fine-tune it with the given dataset. Do required pre-processing on the given dataset. Also, you will be developing a web application that integrates the fine-tuned model and recognize the class of the text provided by the user.

Requirements:

Training Module:

* Pre-process the given dataset.

* Fine tune the pre-trained model with given dataset.  

Users should be able to do the following in the Web App:

* Able to enter a text input.

* Pre-process the text for the deep learning model.

* Predict the class of the text using the fine tuned model.

* View the predicted class label and probability.

Justify the selection of pre-trained model, model architecture and Pre-process on the provided dataset to the Hackthon panel.

Good to Have

Dockerize the solution and deploy using docker-compose

Submission:

• Submit the source code for the training and web application.
• Submit a brief report explaining your model selection justification.

Assignment 2:

Resolve the bug in torch code