This is a TensorFlow implementation of Fine-Grained Head Pose Estimation Without Keypoints .

1. You need a CUDA-compatible GPU to train the model.
2. You should first download 300W-LP for head pose estimation.

• TensorFlow 1.4.1
• TF-Slim
• Python3.6
• Ubuntu 16.04
• Cuda 8.0

1. Setup_data
2. Training
3. Demo
4. Models

Use the script python utils/data_preprocess.py to generate your own annotation file and face images from 300W-LP dataset. One row for one image;
Row format: image_file_path pitch raw roll;
Here is an example:

LFPW_image_test_0236_16_1.jpg,-83.1405049927369,-12.156124797978574,-2.719337151759615
LFPW_image_train_0095_5_2.jpg,-31.33569213834624,-8.705391993430364,19.86217644601874
...

Or you can use my annotation files 300W_LP_headpose_anno.txt under the folder data/ directly.

• Download pre-trained models and weights of backbones.The current code supports VGG16/ResNet_V1/MobileNet_Series models.

• Pre-trained models are provided by slim, you can get the pre-trained models from Google Driver or BaiduYun Driver. Uzip and place them in the folder data/imagenet_weights. For example, for VGG16 model, you also can set up like:

  mkdir -p data/imagenet_weights
  cd data/imagenet_weights
  wget -v http://download.tensorflow.org/models/vgg_16_2016_08_28.tar.gz
  tar -xzvf vgg_16_2016_08_28.tar.gz
  mv vgg_16.ckpt vgg16.ckpt
  cd ../..

  For ResNet101, you can set up like:

  mkdir -p data/imagenet_weights
  cd data/imagenet_weights
  wget -v http://download.tensorflow.org/models/resnet_v1_101_2016_08_28.tar.gz
  tar -xzvf resnet_v1_101_2016_08_28.tar.gz
  mv resnet_v1_101.ckpt res101.ckpt
  cd ../..

• Train Run python trainval_net.py to train the model.

python trainval_net.py \
    --cfg experiment/cfg/mobile.yml \
    --weight data/imagenet_weights/mobile.ckpt \
    --gpu_id 0 \
    --data_dir /home/oeasy/Downloads/dataset/head_pose/300W_LP_headpose \
    --annotation_path /home/oeasy/PycharmProjects/deep-head-pose-tf/300W_LP_headpose_anno.txt \
    --net mobile

Please see details in the script trainval_net.py. By default, trained networks are saved under:

output/[NET]/[DATASET]/default/

Test outputs are saved under:

output/[NET]/[DATASET]/default/[SNAPSHOT]/

Tensorboard information for train and validation is saved under:

tensorboard/[NET]/[DATASET]/default/
tensorboard/[NET]/[DATASET]/default_val/

• For ckpt demo Download trained models from Models, then uzip to the folder output/, modify your path of trained model run python tools/demo.py directly.

• For frozen graph inference Download the pb models(contained in Models) or frozen your model by yourself using script tools/convert_ckpt_to_pb.py, modify your path of trained model, then run python tools/inference.py.

Under updating...