Assignment 5 - Course Project

For the course project, you will pick any dataset of your choice and apply the concepts learned in this course to train deep learning models end-to-end with PyTorch, experimenting with different hyperparameters & metrics. You will prepare a project report in the form a blog post to summarizing your results & findings (the blog post is mandatory).

Submission Link: https://forms.gle/rHxCqBhDGbh2Zn8L8

Deadline: Sat, Jul 2, 9 PM IST

Starter Notebook: There is no starter notebook for the project. Please use the “New Notebook” option from your Jovian profile to create a new notebook.

Guidelines for completing the project

  1. Find a dataset online (either download and existing dataset or create one from web scraping, Google images etc.)
  2. Understand and describe the modeling objective clearly
    1. What type of data is it? (images, text, audio etc.)
    2. What type of problem is it? (regression, classification, generative modeling, etc.)
  3. Exploratory data analysis - explore the data by plotting graphs and answer any questions you may have
  4. Modeling - try 4-5 approaches
    1. Define a model (network architecture)
    2. Pick some hyperparameters
    3. Train the model
    4. Make predictions on samples
    5. Evaluate on test dataset
    6. Save the model weights
    7. Record the metrics
  5. Conclusions - summarize your learning & identify opportunities for future work
  6. Write a blog post to describe your experiments and summarize your work - this is the final submission. Use Github pages or Medium. Submit here: https://forms.gle/rHxCqBhDGbh2Zn8L8

Notice:
Given everyone needs a bit more time to work on the course project, we are making the blog post optional for the 2nd July deadline; Just focus on completing the course project, you can work on the blog post submission and share with us before the graduation day.

9 Likes

Hello colleagues,

Have anyone started working on the Course Project? Share your ideas here.

2 Likes

Hi,
Planning to use dataset from HackerEarth: https://www.hackerearth.com/challenges/competitive/hackerearth-deep-learning-challenge-auto-tag-images-gala/problems/

4 Likes

is it possible to use pretrained models like vgg

Yes. Sure. You can use any model architecture, any dataset, any loss function, any accuracy score, etc.

This assignment is basically something that you need to do on your own from end to end.

1 Like

I am thinking of doing project on DCGAN.

2 Likes

Does anybody have any good resources for finding interesting and unique datasets?

Can we get an article or post on Jovian that is a dataset resource compilation?

Thank you.

5 Likes

I just finished getting the data ready, decided on the model architecture and then i commited the notebook to jovian. But now when i try to run the notebook from jovain on kaggle the notebook doens’t show the cells on kaggle, Only the first cell is visible. How am i supposed to continue the project now. Do i have to re-do it from the beginning ?

Try restarting the kernel

Greetings !

I just started to work on this assignment .

Here is my draft :

Any suggestions are appreciated …

Thank You !

1 Like

Nothing work… I had to re-do it from the beginning lost all progress :upside_down_face:

Can we use the dataset from the kaggle competition? It would be good just to focus on a project and then share it on a blogpost if we had good results!

6 Likes

yes of course you can use

Hi guys, I am trying to build my course project around this dataset I found on kaggle
However, I am having problems to download the data, because the images and the labels are contained in two separate folders. I am trying to build a custom Dataset using the dataset class from pytorch, but I am not able to join the images with their respective names.
this is the link to my notebook https://www.kaggle.com/micheledifazio/flowers-cnn
and this is the code I am using:
class MyDataset(Dataset):

def __init__(self, csv_file, root_dir, transform=None):
    self.labels = pd.read_csv(csv_file)
    self.root_dir = root_dir
    self.transform = transform
    
def __len__(self):
    return len(self.labels)

def __getitem__(self, index):
    **image_name = os.path.join(self.root_dir, self.labels.iloc[index, 0])**
    image = io.imread(image_name)
    label = torch.tensor(self.labels.iloc[index, 1])
    
    if self.transform:
        image = self.transform(image)
        
    return (image, label)

I think that the problem is in the variable img_name, but I do not know how to fix it. Can you help me? Thanks a lot in advance :kissing_heart:

Deadline: Sat, Jul 1, 9 PM IST

When is the deadline exactly? July 1st lands on Wednesday but in the quote above, it says the due date is Saturday.

4 Likes

Could we have details on due date?
There is no Saturday 1rst July, should we assume and plan accordingly for Saturday 4th July ?

1 Like

A post was split to a new topic: How to resolve CUDA Out of Memory error?

Hey folks! I am working on the project on kaggle and locally and I included some maps there made with folium package. However, when commited to jovian it is not showing up . The precise output is as follows:
Make this Notebook Trusted to load map: File -> Trust Notebook
Is there any way I can include it without making a picture or snapshot? Thanks in advance!

What is the minimum score we need to obtain in the datascience competition to pass?

It should above the baseline 16% (0.16748).

1 Like