31434116 / grlmodel

code for IJCAI 18 paper: Grouping Attribute Recognition for Pedestrian with Joint Recurrent Learning

code for IJCAI 18 paper: Grouping Attribute Recognition for Pedestrian with Joint Recurrent Learning

runtime environment

• python version : 3.4
• tensorflow version: >= 1.4

resource

best rap model is: 链接: https://pan.baidu.com/s/1m_bOW2x9_fyhy8dSlPWCuw 提取码: y9ip 
best peta model is: 链接: https://pan.baidu.com/s/1acNTb668IVppJyhfzop_ow 提取码: 3upe
rap TFRecord file: 链接: https://pan.baidu.com/s/15MRSqdg7izo8HgnOD2uCTQ 提取码: 85uk 
dataset label file: 链接: https://pan.baidu.com/s/16D1spNxN2SjZZR5l6v22Uw 提取码: iswu

prepare data (use rap as example)

• ROI data:

use pose estimation model provided in Spindle Net (github link is https://github.com/yokattame/SpindleNet) to get region proposal data.

• put rap or peta label data and region proposal data together.

file format is:

# 0 
CAM12_2014-03-05_20140305110334-20140305111754_tarid1199_frame8675_line1.png
-1  -1  -1  1  -1  1  -1  1  -1  -1  -1  1  -1  -1  -1  1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  1  -1  -1  1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1  -1
19.75390625 0 68 48.6875
0 28.796875 68 112.9609375
0 72.4375 68 188.8125
49.3738839286 29.6875 68 112.8125
0 31.3203125 31.921875 97.8203125
20.4285714286 85.796875 47.5714285714 186.5859375
8.18247767857 74.8125 35.3253348214 174.5625

that is

image index
attribute labels
head region coordinate
up region coordinate
down region coordinate
left arm coordinate
right arm coordinate
left leg coordinate
eight leg coordinate

• run data/build_rap_region_data.py to get tensorflow TFRecord input

train

• install bazel
• in root dir, run "bazel build //inception:rap_train" or "bazel build //inception:rap_test"

run command for example:

TRAIN_DIR=DIR for train models
RAP_DATA_DIR= DIR for rap tensorflow TFRecord file
MODEL_PATH=DIR for pretrained model

bazel-bin/inception/rap_train \
  --train_dir="${TRAIN_DIR}" \
  --data_dir="${RAP_DATA_DIR}" \
  --pretrained_model_checkpoint_path="${MODEL_PATH}" \
  --fine_tune=False \
  --initial_learning_rate=0.1 \
  --input_queue_memory_factor=1 \
  --num_gpus=1 \
  --max_steps=1001

test

TRAIN_DIR=DIR for train models
EVAL_DIR= DIR for eval event logs
RAP_DATA_DIR=DIR for rap tensorflow TFRecord file

    
  bazel-bin/inception/rap_eval \
  --eval_dir="${EVAL_DIR}" \
  --data_dir="${RAP_DATA_DIR}" \
  --subset=validation \
  --num_examples=8317 \
  --checkpoint_dir="${TRAIN_DIR}" \
  --input_queue_memory_factor=1 \
  --run_once