A Compact Embedding for Facial Expression

A_Compact_Embedding_for_Facial_Expression_Similarity(CVPR2019)

Note!

We find that densenet block is not import for Model. you can easily replace it with Conv3*3
Our model structure is the same as maskrcnn-benchmark

Prerequisites

PyTorch 1.0
- pip3 install torch torchvision
Easydict
- pip3 install easydict
Apex
Ninja
- sudo apt-get install ninja-build
tqdm
- pip3 install tqdm

Model Zoo

Supported Model

ResNet18
Inception N22
Inception Resnet
ResNet50

Datasets

Be Careful! Our model does not use all the triple pairs because of the loss of crawler data. In addition, we use our own align model. If you need our processed data, please email me qiulingteng@stu.hit.edu.cn. google original dataset is in here.

For the benefit of everyone, hence I upload my processed data to Baidu WebDisk so that everyone can download and research.
Note that our collected have a lit bit loss due to orignial datasets website link error. link address is given by：

Face_Emotion_Data

The format of annotations you can refer to train_list and test_list.
Key: ndcm

File Structure

  -- PROJECT_ROOT
  -- Triple Dataset
     |-- train_align
     |-- annotations

In annotations, your training triple data name save into *.txt
All of it, you can change in paths_catalog.py

Link

we must build the env for training

make link

soft link to result

Experiment Results

Backbone	ONE_CLASS ACC	TWO_CLASS ACC	THREE_CLASS ACC	ACC
InceptionNN2+denseblock	76.4	78.8	77.8	77.3
InceptionResnet+denseblock	64.1	70.0	65.2	66.9
Resnet18+Conv3*3	78.2	80.5	80.2	79.7
Resnet50+Conv3*3	78.5	81.3	80.9	80.3

Training

create the config file of dataset:train_list.txt
modify the config/*.yaml according to your requirements

Distributed Training

We use the official torch.distributed.launch in order to launch multi-gpu training. This utility function from PyTorch spawns as many Python processes as the number of GPUs we want to use, and each Python process will only use a single GPU.

For each experiment, you can just run this script:

export NGPUS=8
python -m torch.distributed.launch --nproc_per_node=$NGPUS train.py

Non-distributed Training

The above performance are all conducted based on the non-distributed training. For each experiment, you can just run this script:

bash train.sh

Inference

In the evaluator, we have implemented the multi-gpu inference base on the multi-process. In the inference phase, the function will spawns as many Python processes as the number of GPUs we want to use, and each Python process will handle a subset of the whole evaluation dataset on a single GPU.

input arguments in shell:
```
bash inference.sh
```

debanga/Facial_Expression_Similarity