Convolutional Neural Networks Training

Infoshare is the largest tech community in CEE and the organizer of the leading tech conference in Gdańsk. It connects startups, investors, corporations, and innovation enthusiasts. It promotes entrepreneurship, knowledge sharing, and networking. Through events, competitions, and programs, it supports the development of the tech ecosystem in Poland and the region.

An intensive, practical training course on convolutional neural networks (CNNs). During this two-day course, participants will gain in-depth theoretical knowledge and, most importantly, practical skills in designing, implementing, and training convolutional neural networks. The course combines theory with practice, emphasizing hands-on experience in an 80% workshop to 20% lecture ratio.

• Programmers, data scientists, machine learning engineers, and researchers who want to deepen their knowledge of CNNs and gain practical skills in their implementation and optimization.
• Individuals with a basic knowledge of Python and foundational knowledge of machine learning and neural networks.
• For those with minimal experience with ML libraries such as scikit-learn, TensorFlow, or PyTorch.
• Individuals working in a Jupyter Notebook environment.

• Designing and implementing advanced CNN architectures.
• Techniques for optimizing and fine-tuning convolutional models.
• Practical application of transfer learning in vision tasks.
• Implementing and optimizing CNN models in real-world projects.

• Day 1

  • Introduction to neural networks and convolutional networks

  • Basics of neural network architecture

  • CNN architecture

  • Comparison of CNNs with traditional neural networks

  • Convolutional and pooling layers

  • Implementing convolutional layers in PyTorch

  • Designing and optimizing pooling layers

  • Workshop: building a simple CNN

    • Creating a CNN model from scratch

    • Analyzing the impact of different architectures on performance

  • Transfer learning techniques in CNNs

    • Using pre-trained models

    • Fine-tuning models on your own data

• Day 2

  • Advanced CNN architectures

    • Implementing ResNet and Inception

    • Comparative analysis of the performance of different architectures

  • Optimization and regulation of CNNs

    • Regularization techniques: dropout, batch normalization

    • Hyperparameter optimization strategies

  • Workshop: solving complex vision problems

    • Implementing a model for image classification

    • Creating an object detection system

  • Implementing CNN models in practice

    • Optimizing models for performance

    • Integrating CNNs with real-time applications

16 h/2 days