Skip to content

Commit

Permalink
Merge pull request #10 from Idein/update-software
Browse files Browse the repository at this point in the history
Update software
  • Loading branch information
terasakisatoshi committed Dec 21, 2018
2 parents 9ff89c5 + bb3e5a2 commit 835497d
Show file tree
Hide file tree
Showing 18 changed files with 786 additions and 171 deletions.
54 changes: 41 additions & 13 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# chainer-pose-proposal-net

- This is an implementation of [Pose Proposal Networks](http://openaccess.thecvf.com/content_ECCV_2018/papers/Sekii_Pose_Proposal_Networks_ECCV_2018_paper.pdf) with Chainer including training and prediction tools.
- This is an (unofficial) implementation of [Pose Proposal Networks](http://openaccess.thecvf.com/content_ECCV_2018/papers/Sekii_Pose_Proposal_Networks_ECCV_2018_paper.pdf) with Chainer including training and prediction tools.

# License

Expand All @@ -19,31 +19,43 @@ This project is licensed under the terms of the [license](LICENSE).
### MPII

- If you train with COCO dataset you can skip.
- Access [MPII Human Pose Dataset](http://human-pose.mpi-inf.mpg.de/) and jump to `Download` page. Then download and extract both `Images (12.9 GB)` and `Annotations (12.5 MB)`.
- Access [MPII Human Pose Dataset](http://human-pose.mpi-inf.mpg.de/) and jump to `Download` page. Then download and extract both `Images (12.9 GB)` and `Annotations (12.5 MB)` at `~/work/dataset/mpii_dataset` for example.

#### Create `mpii.json`

We need decode `mpii_human_pose_v1_u12_1.mat` to generate `mpii.json`. This will be used on training or evaluating test dataset of MPII.

```
$ sudo docker run --rm -v $(pwd):/work -v path/to/dataset:/data -w /work idein/chainer:4.5.0 python3 convert_mpii_dataset.py /data/mpii_human_pose_v1_u12_2/mpii_human_pose_v1_u12_1.mat /data/mpii.json
$ sudo docker run --rm -v $(pwd):/work -v path/to/dataset:mpii_dataset -w /work idein/chainer:4.5.0 python3 convert_mpii_dataset.py mpii_dataset/mpii_human_pose_v1_u12_2/mpii_human_pose_v1_u12_1.mat mpii_dataset/mpii.json
```

It will generate `mpii.json` at `path/to/dataset` where is the root directory of MPII dataset. For those who hesitate to use Docker, you may edit `config.ini` as necessary.
It will generate `mpii.json` at `path/to/dataset`. Where `path/to/dataset` is the root directory of MPII dataset, for example, `~/work/dataset/mpii_dataset`. For those who hesitate to use Docker, you may edit `config.ini` as necessary.

### COCO

- If you train with MPII dataset you can skip.
- Access [COCO dataset](http://cocodataset.org/) and jump to `Dataset` -> `download`. Then download and extract `2017 Train images [118K/18GB]`, `2017 Val images [5K/1GB]` and `2017 Train/Val annotations [241MB]`.
- Access [COCO dataset](http://cocodataset.org/) and jump to `Dataset` -> `download`. Then download and extract `2017 Train images [118K/18GB]`, `2017 Val images [5K/1GB]` and `2017 Train/Val annotations [241MB]` at `~/work/dataset/coco_dataset:/coco_dataset` for example.

## Running Training Scripts

OK let's begin!

```
$ sudo docker run --rm -v $(pwd):/work -v path/to/dataset:/data -w /work idein/chainer:4.5.0 python3 train.py
$ cat begin_train.sh
cat config.ini
docker run --rm \
-v $(pwd):/work \
-v ~/work/dataset/mpii_dataset:/mpii_dataset \
-v ~/work/dataset/coco_dataset:/coco_dataset \
--name ppn_idein \
-w /work \
idein/chainer:5.1.0 \
python3 train.py
$ sudo bash begin_train.sh
```


- Optional argument `--runtime=nvidia` maybe require for some environment.
- This will train a model the base network is MobileNetV2 with MPII dataset located in `path/to/dataset` on host machine.
- It will train a model the base network is MobileNetV2 with MPII dataset located in `path/to/dataset` on host machine.
- If we would like to train with COCO dataset, edit a part of `config.ini` as follow:

before
Expand Down Expand Up @@ -81,10 +93,22 @@ model_name = resnet18

# Prediction

- Very easy, all we have to do is:
- Very easy, all we have to do is, for example:

```
$ sudo bash run_predict.sh ./trained
```

- If you would like to configure parameter or hide bounding box, edit a part of `config.ini` as follow:

```
$ sudo docker run --rm -v $(pwd):/work -v path/to/dataset:/data -w /work idein/chainer:4.5.0 python3 predict.py
[predict]
# If `False` is set, hide bbox of annotation other than human instance.
visbbox = True
# detection_thresh
detection_thresh = 0.15
# ignore human its num of keypoints is less than min_num_keypoints
min_num_keypoints= 1
```

# Demo: Realtime Pose Estimation
Expand All @@ -102,24 +126,28 @@ docker build -t ppn .
$ sudo bash build.sh
```

Here is an result of ResNet18 trained with COCO running on laptop PC.

![](readmedata/cpu-example.gif)

## Run video.py

- Set your USB camera that can recognize from OpenCV.

- Run `video.py`

```
$ python video.py
$ python video.py ./trained
```

or

```
$ sudo bash run_video.sh
$ sudo bash run_video.sh ./trained
```

## High Performance Version
- To use feature of [Static Subgraph Optimizations](http://docs.chainer.org/en/stable/reference/static_graph_design.html) to accelerate inference speed, we should install Chainer 5.0.0 and CuPy 5.0.0 .
- To use feature of [Static Subgraph Optimizations](http://docs.chainer.org/en/stable/reference/static_graph_design.html) to accelerate inference speed, we should install Chainer 5.y.z and CuPy 5.y.z e.g. 5.0.0 or 5.1.0 .
- Prepare high performance USB camera so that takes more than 60 FPS.
- Run `high_speed.py` instead of `video.py`
- Do not fall from the chair with surprise :D.
Expand Down
Loading

0 comments on commit 835497d

Please sign in to comment.