Pytorch Tutorial - Simple Object Detector

Repo for new students willing to learn Pytorch. It implements a simple object detector based on VGG16 with two braches -- probability of the object being in the image and its bounding box. Images must be of size 224x224x3.

Prerequisites

To understand the repo I recommend first doing the official Pytorch Beginner Tutorials.

Dependencies

Apart from PyTorch and Tensorboard:

pip install torchvision  # pretrained VGG16 parameters
pip install matplotlib   # for visualization
pip install tensorboardX # for visualization

Train/Test Example

Train:

bash launch/run_train.sh

Test:

bash launch/run_test.sh

For any doubt contact albert.pumarola.peris at gmail.com