MSPN

This implements training of WSPN on the COCO2014 dataset, mainly modified from [github][9].

Repo Structure

This repo is organized as following:

$MSPN_HOME
|-- cvpack
|
|-- dataset
|   |-- COCO
|   |   |-- det_json
|   |   |-- gt_json
|   |   |-- images
|   |       |-- train2014
|   |       |-- val2014
|   |
|   |-- MPII
|       |-- det_json
|       |-- gt_json
|       |-- images
|   
|-- lib
|   |-- models
|   |-- utils
|
|-- exps
|   |-- mspn.2xstg.coco
|   
|---test
|	|--env_npu.sh
|	|--train_full_8p.sh
|	|--train_performance_1p.sh
|	|--train_performance_8p.sh
|
|-- model_logs
|
|-- README.md
|-- requirements.txt

Quick Start

Installation

1. Install Pytorch referring to [Pytorch website][2].

2. Clone this repo, and config MSPN_HOME in /etc/profile or ~/.bashrc, e.g.

export MSPN_HOME='/path/of/your/cloned/repo'
export PYTHONPATH=$PYTHONPATH:$MSPN_HOME

3. Install requirements:

pip3 install -r requirements.txt

4. Install COCOAPI referring to [cocoapi website][3], or:

git clone https://github.com/cocodataset/cocoapi.git $MSPN_HOME/lib/COCOAPI
cd $MSPN_HOME/lib/COCOAPI/PythonAPI
make install

Dataset

COCO

1. Download images from [COCO website][4], and put train2014/val2014 splits into $MSPN_HOME/dataset/COCO/images/ respectively.

2. Download ground truth from [Google Drive][6], and put it into $MSPN_HOME/dataset/COCO/gt_json/.

3. Download detection result from [Google Drive][6], and put it into $MSPN_HOME/dataset/COCO/det_json/.

Model

Download ImageNet pretained ResNet-50 model from [Google Drive][6], and put it into $MSPN_HOME/lib/models/. For your convenience, We also provide a well-trained 2-stage MSPN model for COCO.

Log

Create a directory to save logs and models:

mkdir $MSPN_HOME/model_logs

Train

Go to specified experiment repository, e.g.

# training 1p performance
bash ./test/train_performance_1p.sh 

# training 8p accuracy
bash ./test/train_full_8p.sh 

# training 8p performance
bash ./test/train_performance_8p.sh 

Test

python -m torch.distributed.launch --nproc_per_node=gpu_num test.py -i iter_num

the gpu_num is the number of gpus, and iter_num is the iteration number you want to test. Remenber that test in gpu environment.

MSPN training result

Results on COCO val dataset

Acc@1	FPS	Npu_nums	Epochs	AMP_type
-	64.727	1	4	O1
74.5	341.4	8	4	O1

Statement

For details about the public address of the code in this repository, you can get from the file public_address_statement.md