Project Setup

1. Environment Setup

Create a virtual environment and install the required dependencies:

pip install -r requirements.txt

2. Dataset

Images

Download the COCO2017 image dataset from this link.

Annotations

Download the LVIS annotation files from this link.

Update constants

Create a python file constants.py and paste the following constants with updated paths.

COCO2017_TRAIN_PATH = "<path_to_train_dataset_folder>"
COCO2017_VAL_PATH   = "<path_to_val_dataset_folder>"
COCO2017_TEST_PATH  = "<path_to_test_dataset_folder>"
TRAIN_ANNOT_PATH = "<path_to_train_annotations_file>"
VAL_ANNOT_PATH   = "<path_to_val_annotations_file>"

# Useful ??
TEST_ANNOT_PATH  = "<path_to_test_annotations_file>"

Resources

TorchVision Object Detection Finetuning Tutorial
LVIS Data Format