tracking-with-rnn

Videos and codebase of First step toward model-free, anonymous object tracking with recurrent networks.

Requirements

All models require Theano to train & test.

Additionally, ConvTracker requires keras.

Files

conv_base.py - ConvTracker recurrent_base.py - RecTracker-ID recurrent_att.py - RecTracker-Att-N. N can be specified in --grid_size option. data_handler.py - Moving MNIST dataset generator

Usage

Run train.sh to train the RecTracker-Att-1 model.

Run test.sh to test the RecTracker-Att-1 model under different configurations.

RecTracker-ID and ConvTracker accept the same options.

Name		Name	Last commit message	Last commit date
Latest commit History 19 Commits
.gitignore		.gitignore
ConvTracker trained on single MNIST-1.mp4		ConvTracker trained on single MNIST-1.mp4
README.md		README.md
RecTracker-Att-1 distracted away in two moving objects.mp4		RecTracker-Att-1 distracted away in two moving objects.mp4
RecTracker-Att-1 on single object with attention.mp4		RecTracker-Att-1 on single object with attention.mp4
RecTracker-Att-1 with Mask displayed.mp4		RecTracker-Att-1 with Mask displayed.mp4
RecTracker-Att-3 with mask displayed.mp4		RecTracker-Att-3 with mask displayed.mp4
RecTrackerID trained on single MNIST-1.mp4		RecTrackerID trained on single MNIST-1.mp4
RecTrackerID trained on single but tested on multiple.mp4		RecTrackerID trained on single but tested on multiple.mp4
conv_base.py		conv_base.py
data_handler.py		data_handler.py
mnist.h5		mnist.h5
recurrent_att.py		recurrent_att.py
recurrent_att_test.py		recurrent_att_test.py
recurrent_base.py		recurrent_base.py
recurrent_local_online.log		recurrent_local_online.log
recurrent_local_online.py		recurrent_local_online.py
recurrent_local_online_profile.log		recurrent_local_online_profile.log
test.sh		test.sh
train.sh		train.sh

BarclayII/tracking-with-rnn

Folders and files

Latest commit

History

Repository files navigation

tracking-with-rnn

Requirements

Files

Usage

About

Resources

Stars

Watchers

Forks

Languages