Skip to content

Latest commit

 

History

History
41 lines (30 loc) · 2.05 KB

README.md

File metadata and controls

41 lines (30 loc) · 2.05 KB

VeryDeepLearning

  1. Assignment 1 - Assignment 1 is regarding Forest Cover type dataset. The dataset includes 7 different forest cover types. Need to predict the forest cover type based on input features provided. It is solved using Fully Connected Neural Network.

  2. Assignment 2 - This Assignment contains 2 tasks,

    • Task 1: It is a classification task on CIFAR10 dataset. Need to classify dataset with 10 different categories. It is solved using Convolutional Neural Networks.
    • Task 2: It is again a classification task from Kaggle, Dog Breed Classification Dataset. For this problem we have been usiing Transfer Learning approach. I tried it using Alexnet, Resnet18, pretrained networks. Further tried to improve accuracy by adding more layers at the end. received a score of 1.4 in kaggle.

  3. Assignment 3 - This Assignment contains 2 tasks,

    • Task 1: Sementic Segmentation

      1. Download and extract the bags dataset:
      2. Write a pytorch Dataset class which can iterate over the dataset
      3. Train a fully-convolutional network on the training dataset
      4. Report the mean IoU score after each epoch for both training dataset and validation dataset.
      5. Visualize your network output on one image of the validation set

    • Task 2: Visualizing Convolutional Neural Networks 1.Download a pretrained AlexNet model and plot the weights of the First convolutional layer.

  4. Assignment 4 - Sentiment Analysis Using RNN and LSTM

    • In this exercise, you will understand and implement the below concepts :

    • Reading a text dataset.

    • Pre-processing the text such as Stopwords Removal etc.

    • Word embedding

      1. Vocabulary Indexing
      2. Glove embedding

    • Implementing Dataloaders

    • A Neural Network

      1. RNNs
      2. LSTMs

    • Evaluation of the network

    The task here is to develop a simple sentiment analysis model for a dataset by understanding if the review is positive or negative for each statement.