Requirements

• Windows with an NVIDIA GPU (worked and tested on NVIDIA RTX 3500 Ada Lovelace)
• Python 3.10.x
• NVIDIA driver compatible with CUDA 11.2
• CUDA Toolkit 11.2 and cuDNN 8.1 (required for TensorFlow 2.10 on Windows)
• Train and test datasets: https://drive.google.com/drive/folders/1Xn9ELRoW3GmHMUgrQLBTzzYO4INvSYIw

Setup (Conda)

Create and activate a virtual environment (recommended):

conda env create -f environment.yaml
conda activate deep-learning

This will install Python 3.10 and all required packages and dependencies to run TensorFlow 2.10 with GPU support on Windows.

Notes

• TensorFlow 2.10 requires CUDA 11.2 and cuDNN 8.1 on Windows; other versions can lead to import/runtime errors.
• The provided requirements.txt pins TensorFlow to < 2.11 to stay compatible with Python 3.10 and the CUDA/cuDNN versions above.