whateverhack-burger-task2
Code for the 1st place solution in task 2 whateverhack
MIT License
Stars
11
Ecosystems:
Jupyter Notebook
WhateverHack Task 2
Code for the 1st place solution in task 2 whateverhack. Task description and leaderboard.
We used Mask R-CNN to detect furniture in photos of real estate. To achieve best results we use an ensemble prediction of several different model checkpoints. Base model implementation matterport/Mask_RCNN.
Our team:
- Renat Bashirov (linkedin, kaggle, github)
- Ruslan Baikulov (linkedin, kaggle, github)
- Ali Aliev (linkedin, kaggle, github)
Predictions
Requirements
Hardware:
The solution was obtained on a PC with the following hardware:
- 64 GB of RAM
- 2x Nvidia GTX 1080Ti
Software:
- Linux Ubuntu 16.04
- Nvidia drivers, CUDA >= 9, cuDNN >= 7
- Docker, nvidia-docker
The provided dockerfile is supplied to build image with cuda support and cudnn.
How to run
The solution was written under hackathon conditions, therefore in the current state it is rather difficult to reproduce.