dl-pose-hrn

Deep High-Resolution Network

This container is used to estimate pose. Pose estimation using this container is a two step process.

1. Use Yolo to detect the humans in an image.
2. Use this container to estimate the pose of each human in an image.

To build the container, you will need to download the models first; see Deep High-Resolution Network instructions. After you download the models, you should place the models in a directory called models with the following structure.

models/
└── pytorch
    ├── imagenet
    ├── pose_coco
    └── pose_mpii

Then use the build script.

./build.sh

You need to set up your input directories in a certain way. Look here for more information on the meaning of this directory layout.

custom/
├── annots
├── cuts
├── images
└── inspect

Run the Yolo container to detect where the humans are located in your images.

docker run -it \
    -v `pwd`/custom:/yolo/custom \
    --gpus=all \
    --shm-size=5g \
    oneoffcoder/dl-pose-yolo \
      --image_folder custom/images \
      --annot_folder custom/annots \
      --inspect_folder custom/inspect \
      --cut_folder custom/cuts \
      --batch_size 8 \
      --n_cpu 8

Now, run this container to estimate the human poses.

docker run -it \
    -v `pwd`/custom:/hrn/custom \
    --gpus=all \
    --shm-size=5g \
    oneoffcoder/dl-pose-hrn \
      --circle 0 \
      --label 0 \
      --line 1 \
      --line_thickness 2 \
      --cfg custom/w32_256x256_adam_lr1e-3.yaml

The key is in the configuration file, specified by the --cfg flag, custom/w32_256x256_adam_lr1e-3.yaml. Note that this file should be placed at the root (under custom). Pay attention to the ANNOTS, IMAGES, CUTS and FINAL parameters for DATASET. Here's the example.

AUTO_RESUME: true
CUDNN:
  BENCHMARK: true
  DETERMINISTIC: false
  ENABLED: true
DATA_DIR: ''
GPUS: (0,)
OUTPUT_DIR: 'custom/output'
LOG_DIR: 'log'
WORKERS: 24
PRINT_FREQ: 100

DATASET:
  COLOR_RGB: true
  DATASET: custom
  DATA_FORMAT: jpg
  FLIP: true
  NUM_JOINTS_HALF_BODY: 8
  PROB_HALF_BODY: -1.0
  ROOT: 'data/example/'
  ROT_FACTOR: 30
  SCALE_FACTOR: 0.25
  TEST_SET: valid
  TRAIN_SET: train
  ANNOTS: 'custom/annots'
  IMAGES: 'custom/images'
  CUTS: 'custom/cuts'
  FINAL: 'custom/final'
MODEL:
  INIT_WEIGHTS: true
  NAME: pose_hrnet
  NUM_JOINTS: 16
  PRETRAINED: 'models/pytorch/pose_mpii/pose_hrnet_w32_256x256.pth'
  TARGET_TYPE: gaussian
  IMAGE_SIZE:
  - 256
  - 256
  HEATMAP_SIZE:
  - 64
  - 64
  SIGMA: 2
  EXTRA:
    PRETRAINED_LAYERS:
    - 'conv1'
    - 'bn1'
    - 'conv2'
    - 'bn2'
    - 'layer1'
    - 'transition1'
    - 'stage2'
    - 'transition2'
    - 'stage3'
    - 'transition3'
    - 'stage4'
    FINAL_CONV_KERNEL: 1
    STAGE2:
      NUM_MODULES: 1
      NUM_BRANCHES: 2
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      NUM_CHANNELS:
      - 32
      - 64
      FUSE_METHOD: SUM
    STAGE3:
      NUM_MODULES: 4
      NUM_BRANCHES: 3
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      - 4
      NUM_CHANNELS:
      - 32
      - 64
      - 128
      FUSE_METHOD: SUM
    STAGE4:
      NUM_MODULES: 3
      NUM_BRANCHES: 4
      BLOCK: BASIC
      NUM_BLOCKS:
      - 4
      - 4
      - 4
      - 4
      NUM_CHANNELS:
      - 32
      - 64
      - 128
      - 256
      FUSE_METHOD: SUM
LOSS:
  USE_TARGET_WEIGHT: true
TRAIN:
  BATCH_SIZE_PER_GPU: 32
  SHUFFLE: true
  BEGIN_EPOCH: 0
  END_EPOCH: 210
  OPTIMIZER: adam
  LR: 0.001
  LR_FACTOR: 0.1
  LR_STEP:
  - 170
  - 200
  WD: 0.0001
  GAMMA1: 0.99
  GAMMA2: 0.0
  MOMENTUM: 0.9
  NESTEROV: false
TEST:
  BATCH_SIZE_PER_GPU: 1
  MODEL_FILE: ''
  FLIP_TEST: true
  POST_PROCESS: true
  SHIFT_HEATMAP: true
DEBUG:
  DEBUG: true
  SAVE_BATCH_IMAGES_GT: true
  SAVE_BATCH_IMAGES_PRED: true
  SAVE_HEATMAPS_GT: true
  SAVE_HEATMAPS_PRED: true

The circle, label and line flags indicate if you want the final annotated (output) pictures to have circles, labels or lines. After you run the container against your input data, you should have a custom/final folder with your input images annotated with the pose estimations. There will also be a custom/output folder for debugging purposes.

Citation

@misc{oneoffcoder_dl_pose_hrn_2019, 
title={An executable docker container for Deep High-Resolution Network pose estimation}, 
url={https://github.com/oneoffcoder/docker-containers/tree/master/dl-pose-hrn}, 
journal={GitHub},
author={One-Off Coder}, 
year={2020}, 
month={Jan}}