Implementation of Faster R-CNN For Object Detection On Berkeley Deep Drive dataset

Setup Instructions

• Install CUDA 9.0
• Install NVIDIA Driver 390
• Add CUDA paths to the bashrc file
• Create data directory
• Create bdd_data inside data
• Create ImageSets, JPEGImages, Annotations and results directory inside bdd_data
• Create Main directory inside ImageSets
• Copy all training and validation images from BDD100k images to JPEGImages
• Run create_val_list and create_train_list

For other implementation details, obtaining pretrained models and setup, refer https://github.com/jwyang/faster-rcnn.pytorch

For obtaining the BDD dataset, visit http://bdd-data.berkeley.edu

Get models, results, annotations and detections from https://drive.google.com/drive/folders/1v7qRDaLZqXINDbTPERYI_FNQOv6ukjFJ