CornerNet Implementation in Pytorch-Lightning

CornerNet is a new approach to object detection where we detect an object bounding box as a pair of keypoints (top-left corner and bottom-right corner)

This repo contains implementation of CornerNet in Pytorch-Lightning

Paper

Results

Getting started

Requirements

• pytorch 1.9.1
• kornia
• cornerpool
• pytorch-lightning
• tqdm

Clone this Repo

CornerNet_ROOT=/path/to/clone/CornerNet
git clone https://github.com/Jovian-Dsouza/CornerNet.git $CornerNet_ROOT

Dataset

Downlaod the Pascal voc dataset

Train

python train_pl.py

Evaluate

python evaluate.py

References

zzzxxxttt/pytorch_simple_CornerNet

princeton-vl/CornerNet