Fun Projects

EG4338/EG6338 Machine Learning

Kaggle Challenges

Kaggle is a platform for solving general machine learning problems. It offers a wide range of free datasets and challenges that can help you apply your machine learning skills in real-world scenarios. The following are some suggested Kaggle competitions that are suitable for beginners and can help you get started with practical machine learning tasks.

  • House Prices: Advanced Regression Techniques (Link)
    Use advanced regression techniques to predict house prices.

  • Digit Recognizer (Computer Vision Problem) (Link)
    Use computer vision techniques to recognize handwritten digits.

  • MNIST: Image Classification (Link)
    Classify images of handwritten digits using machine learning.

  • Forest Cover Type Prediction (Link)
    Predict forest cover type based on cartographic data.

  • Image Processing: Facial Keypoints Detection (Link)
    Detect facial keypoints in images using image processing techniques.

  • Natural Language Processing: Word2Vec NLP Tutorial (Link)
    Learn natural language processing techniques using Word2Vec.

Hugging Face Models

Hugging Face is an AI community that hosts a wide range of pre-trained models and datasets for computer vision (CV) and natural language processing (NLP) tasks. It is an excellent resource for exploring and applying machine learning techniques in CV and NLP. Below are some popular Hugging Face projects that you can explore:

  • Transformers (Link)
    Backbone for state-of-the-art natural language processing models.

  • Bidirectional Encoder Representations from Transformers (BERT) (Link)
    A transformer-based model for various NLP tasks, including next sentence prediction.

  • A Distilled Version of BERT (DistilBERT) (Link)
    A smaller, faster, cheaper alternative to BERT.

  • Generative Pre-trained Transformer 2 (GPT-2) (Link)
    A state-of-the-art language model for generating human-like text.

  • Segment Anything Model (SAM) (Link)
    A model for segmenting objects in images.

  • Segment Anything Model 2 (SAM2) (Link)
    An extended version of the SAM model for segmenting objects in videos.