This repo provides our team's solution for the task "Overlap Area Estimation" in the Junction X Hanoi 2023 Hackathon. In this task, we are required to determine the overlapping zone of the cameras in different directions. Check the file VTX_OVA for detailed task description, and the file Submission_Guide for submission requirements. Our team decide to choose the model SuperGlue for matching overlap area among viđeo frames. The model has been pretrained and we simply use it without any training.

1. Clone the repo

   git clone https://github.com/hieunm44/overlap-area-estimation.git
   cd overlap-area-estimation

2. Install necessary packages

   pip install -r requirements.txt

3. Go to this Google drive storage to download the dataset and put them in the folder data_btc.
4. Run the task and write result files

   python3 main.py

   Result files will be generated in the folder Next Gen AI (our team's name), following the required format by the Hackathon organizers. The code runs better if you have GPUs on your system.

• SuperGlue source code: https://github.com/magicleap/SuperGluePretrainedNetwork
• SuperGlue visualizations: https://psarlin.com/superglue/