- This project focuses on building deep learning model training and evaluation pipelines for object detection of pavement road distresses in response to IEEE 2020 Global Road Detection Challenge
- Object detection models used: YOLO, Faster R-CNN
- Computer Vision frameworks used: Pytorch, Tensorflow
- See paper for final research insights & results
Hardware used: Nvidia RTX 3090
- Given the RTX 3090 embeds Ampere architecture, it will only work with Nvidia Driver 450+ versions only.
https://docs.nvidia.com/deploy/cuda-compatibility/index.html - Given we can only work with Nvidia Driver versions 450+, we will require CUDA versions 11.0+
https://docs.nvidia.com/deploy/cuda-compatibility/index.html - Given we can now only work with CUDA versions 11.0+, we will require cuDNN versions 8/0+ https://docs.nvidia.com/deeplearning/cudnn/support-matrix/index.html
- See https://medium.com/@dun.chwong/the-simple-guide-deep-learning-with-rtx-3090-cuda-cudnn-tensorflow-keras-pytorch-e88a2a8249bc for more details
- XML_to_TXT_Annotation_Conversion_Pipeline.ipynb to convert XML annotation files to TXT for YOLOv5 use
- A01 - Load and Augment an Image.ipynb to define augmentations to apply to input images