densepose_setup

GETTING STARTED

Step 1: Get COCO data:

make a directory structure:

coco/
annotations/

train/val annotations

wget http://images.cocodataset.org/annotations/annotations_trainval2014.zip

valminusminival2014, minival2014:

wget https://storage.googleapis.com/coco-dataset/external/PASCAL_VOC.zip

images

cd path/to/coco/
mkdir val2014 && gsutil -m rsync gs://images.cocodataset.org/val2014 val2014
mkdir train2014 && gsutil -m rsync gs://images.cocodataset.org/train2014 train2014

structure:

/path/to/coco/
train2014
val2014
annotations/
instances_train2014.json
instances_val2014.json
instances_valminusminival2014.json
instances_minival2014.json
(not sure if keypoints/captions needed, prob not)

Step 2: Set up DensePose

get densepose

DENSEPOSE=/home/austin/densepose
git clone https://github.com/facebookresearch/densepose $DENSEPOSE

get densepose annotations

cd $DENSEPOSE/DensePoseData
bash get_densepose_uv.sh
bash get_DensePose_COCO.sh
bash get_eval_data.sh

set up docker

cd $DENSEPOSE/docker
docker build -t densepose:c2-cuda9-cudnn7 .

nvidia-docker run --rm -it densepose:c2-cuda9-cudnn7 python2 detectron/tests/test_batch_permutation_op.py

modify docker

enter container

nvidia-docker run -v $DENSEPOSE/DensePoseData:/denseposedata -v /home/austin/coco:/coco -it densepose:c2-cuda9-cudnn7 bash
mv /densepose/DensePoseData /densepose/DensePoseDataLocal
ln -s /denseposedata DensePoseData
ln -s /coco /densepose/detectron/datasets/data/coco
ln -s /densepose/DensePoseData/DensePose_COCO/densepose_coco_2014_minival.json /densepose/detectron/datasets/data/coco/annotations/
ln -s /densepose/DensePoseData/DensePose_COCO/densepose_coco_2014_train.json /densepose/detectron/datasets/data/coco/annotations/
ln -s /densepose/DensePoseData/DensePose_COCO/densepose_coco_2014_valminusminival.json /densepose/detectron/datasets/data/coco/annotations/

Exit container and commit changes

docker commit $(docker ps --last 1 -q) densepose:c2-cuda9-cudnn7-wdata

Part 3: Run inference/testing/training commands

nvidia-docker run --rm -v $DENSEPOSE/DensePoseData:/denseposedata -v /home/austin/coco:/coco -it densepose:c2-cuda9-cudnn7-wdata <inference_or_training_command>

testing

nvidia-docker run --rm -v $DENSEPOSE/DensePoseData:/denseposedata -v /home/austin/coco:/coco -it densepose:c2-cuda9-cudnn7-wdata python2 tools/test_net.py --cfg configs/DensePose_ResNet101_FPN_s1x-e2e.yaml TEST.WEIGHTS https://s3.amazonaws.com/densepose/DensePose_ResNet101_FPN_s1x-e2e.pkl NUM_GPUS 1

inference

Single image inference on demo data

nvidia-docker run --rm -v $DENSEPOSE/DensePoseData:/denseposedata -v /home/austin/coco:/coco -it densepose:c2-cuda9-cudnn7-wdata \ python2 tools/infer_simple.py \ --cfg configs/DensePose_ResNet101_FPN_s1x-e2e.yaml \ --output-dir DensePoseData/infer_out/ \ --image-ext jpg \ --wts https://s3.amazonaws.com/densepose/DensePose_ResNet101_FPN_s1x-e2e.pkl \ DensePoseData/demo_data/demo_im.jpg

Multiple image inference on own data

DENSEPOSE=/home/austin/densepose && nvidia-docker run -v $DENSEPOSE/DensePoseData:/denseposedata -v /home/austin/coco:/coco -v /home/austin/data/densepose_mydata:/densepose_mydata -it -p 8888:8888 --name densepose densepose:c2-cuda9-cudnn7-wdata \ python2 tools/infer_simple.py \ --cfg configs/DensePose_ResNet101_FPN_s1x-e2e.yaml \ --output-dir /densepose_mydata/infer_out/ \ --image-ext jpg \ --wts https://s3.amazonaws.com/densepose/DensePose_ResNet101_FPN_s1x-e2e.pkl \ /densepose_mydata/test_imgs/

Part 4: Visualization of outputs (don't run with --rm if want to visualize in jupyter)

jupyter notebook --ip 0.0.0.0 --no-browser --allow-root
notebooks/DensePose-RCNN-Visualize-Results.ipynb
loc: /densepose_mydata/infer_out

Script

if i make a script, use popd/pushd
use detached mode