Temporal Object Pose Estimation : KAIST 2021SP AI604

Video Object Pose estimation project for KAIST 2021SP AI604 : Computer Vision

With pip:

pip3 install -r requirements.txt # --user

See requirements.txt for this case.

Alternatively, with poetry:

poetry update

See pyproject.toml for this case which includes the specs for the compatibility matrix.

Note that as of 04-15-2021, installation with poetry requires latest changes from master. This is due to the version specification conventions for torch and torchvision (i.e. the +cuXXX modifiers) See poetry#3831.

Requires a modified version of the tfrecord package to handle the following list of complications:

• Loading from GCS (Google Cloud Storage) bucket
• Loading a SequenceExample
• Loading from shard(s)

One way to install the package would be the following:

pip3 install git+https://github.com/yycho0108/tfrecord.git@gcs-seq

For further details, visit yycho0108/tfrecord instead.

Yoonyoung (Jamie) Cho : yoonyoung.cho@kaist.ac.kr Jiyong Ahn : ajy8456@gmail.com