What are some amazing projects in Python Related to AI and ML?
There are so many AI enthusiasts that will just go from one online course to the other without implementing what they have learnt so first of all, I would just like to take this one sentence away from the answer and tell you that I’m genuinely glad you want to do projects and get your hands dirty.
Since I have no clue as to the level of AI you are at, I’m going to list projects covering all levels (beginner, intermediate, advanced)
Alright. Let’s get the list down.
- Music Generation: The name is probably self-explanatory and you may choose any dataset from here: The Lakh MIDI Dataset v0.1 for this project.
- Neural Style Transfer: The idea is to take a normal mundane image and take another stylish image and “transfer the style” of the stylish image to the mundane image. Here’s a classic example — a picture of Hoover Tower at Stanford, in the style of The Starry Night:
Here is a link to a dataset you might find useful: fzliu/style-transfer
You can even take this a notch higher and use your algorithm to make an app out of it. Something like that is a great thing to have on your resume should you pursue future endeavours in this field.
3. Autonomous Driving: If you go to the web, you will find ready-made algorithms like the YOLO models to do this task for you. I encourage you to study and implement those algorithms yourself.
4. AI Colorisation of Images: This is by no means a simple project and you may need guidance on how to go about things so I’m pasting a link for your ready reference: Colorful Image Colorization. You can even take this a step further and find a way to colorise videos. Video is not that hard if you’ve got the image part right. I encourage you to try it, the results will surprise you.
5. Image Captioning: The best dataset I can recommend here is the COCO dataset. Here is the link: Common Objects in Context. The idea is to caption an image shown to the algorithm the way a human would.
6. Face Emotion Recognition: This is a relatively simple problem. You take a face and detect the emotion. A dataset you might find useful: Cohn-Kanade (CK and CK+) database Download Site. I encourage you to, however, go a step further and do this in REAL-TIME.
7. Speech-to-Text: I will direct you to an excellent dataset that can help you out here: openslr.org. This website has a lot of data from different languages. I recommend you do the English translation only and move on to the more interesting projects. Just search that website using keyword “English” by Command + F (Mac) or Ctrl + F (Windows) and choose any dataset you like.
8. Machine Translation: Again, I’d say a self-explanatory title. I will be directing you to a website that has many language pairs. Like German — English, Italian — English etc. I encourage you to take it a level higher and also try to attempt the German — Italian translation. Here’s the link: Tab-delimited Bilingual Sentence Pairs
9. Sentiment Analysis: The twitter dataset is the classic here but here is a dataset for Amazon Reviews that I found way more interesting: https://www.kaggle.com/bittlingm...
10. Automatic Text Summarisation: The idea is to create a summary with the major points of the original document. A dataset you might find useful: DMQA
11. Recommender Systems (Movie): The link to a dataset you’ll find useful: MovieLens. The idea is to take a user and produce the rating he will give on a movie.
12. Video Summarisation: The idea is to train a network to summarise the video you fed it. Dataset: A Large and Diverse Labeled Video Dataset for Video Understanding Research. You could take up a similar but easier problem of Automatic Highlights Generation and then build it up to a summarisation problem. Here is an excellent paper for highlights generation (cricket) : https://cse.iitk.ac.in/users/cs3... . I am unaware of a ready dataset that can help you out but I found a really good implementation of this problem at this website Deepcric — Deep learning and Cricket — Deepytics, maybe they can help you out with their dataset that they’ve hand-engineered.
Lastly, I recommend you to try your hands at this project:
13. Take a game like Pac-Man and do some RL to avoid ghosts and eat most pellets (Deep Reinforcement Learning)
This is a complicated problem and it is from the field of Deep Reinforcement Learning which is the hottest field of research in AI. Before taking up this project, however, please make sure you understand the field and its literature. Here is an excelled tutorial for DRL if you wish to enter it:
Deep Reinforcement Learning Course
Let me know if you need help with any of these projects including the last one and good luck for your future endeavours