Skip to content

Latest commit

 

History

History
184 lines (140 loc) · 7.9 KB

File metadata and controls

184 lines (140 loc) · 7.9 KB

Face Detection

Models that are able to detect faces.

Model Name Complexity (GFLOPs) Size (Mp) AP @ [IoU=0.50:0.95] (%) AP for faces > 64x64 (%) WiderFace Easy (%) WiderFace Medium (%) WiderFace Hard (%) Links GPU_NUM
face-detection-0200 0.82 1.83 16.0 86.743 82.917 76.198 41.443 snapshot, model template 2
face-detection-0202 1.84 1.83 20.3 91.938 89.382 83.919 50.189 snapshot, model template 2
face-detection-0204 2.52 1.83 21.4 92.888 90.453 85.448 52.091 snapshot, model template 4
face-detection-0205 2.94 2.02 21.6 93.566 92.032 86.717 54.055 snapshot, model template 4
face-detection-0206 340.06 63.79 34.2 94.274 94.281 93.207 84.439 snapshot, model template 8
face-detection-0207 1.04 0.81 17.2 88.17 84.406 76.748 43.452 snapshot, model template 1

Average Precision (AP) is defined as an area under the precision/recall curve.

Training pipeline

0. Change a directory in your terminal to object_detection.

cd <training_extensions>/pytorch_toolkit/object_detection

If You have not created virtual environment yet:

./init_venv.sh

Else:

. venv/bin/activate

or if You use conda:

conda activate <environment_name>

1. Select a model template file and instantiate it in some directory.

export MODEL_TEMPLATE=`realpath ./model_templates/face-detection/face-detection-0200/template.yaml`
export WORK_DIR=/tmp/my_model
python ../tools/instantiate_template.py ${MODEL_TEMPLATE} ${WORK_DIR}

2. Collect dataset

Download the WIDER Face and unpack it to the ${DATA_DIR} folder.

export DATA_DIR=${WORK_DIR}/data

3. Prepare annotation

Convert downloaded and extracted annotation to MSCOCO format with face as the only one class.

  • Training annotation

    export TRAIN_ANN_FILE=${DATA_DIR}/instances_train.json
    export TRAIN_IMG_ROOT=${DATA_DIR}
    python ./model_templates/face-detection/tools/wider_to_coco.py \
       ${DATA_DIR}/wider_face_split/wider_face_train_bbx_gt.txt \
       ${DATA_DIR}/WIDER_train/images/ \
       ${TRAIN_ANN_FILE}
  • Validation annotation

    export VAL_ANN_FILE=${DATA_DIR}/instances_val.json
    export VAL_IMG_ROOT=${DATA_DIR}
    python ./model_templates/face-detection/tools/wider_to_coco.py \
       ${DATA_DIR}/wider_face_split/wider_face_val_bbx_gt.txt \
       ${DATA_DIR}/WIDER_val/images/ \
       ${VAL_ANN_FILE}

4. Change current directory to directory where the model template has been instantiated.

cd ${WORK_DIR}

5. Training and Fine-tuning

Try both following variants and select the best one:

  • Training from scratch or pre-trained weights. Only if you have a lot of data, let's say tens of thousands or even more images. This variant assumes long training process starting from big values of learning rate and eventually decreasing it according to a training schedule.

  • Fine-tuning from pre-trained weights. If the dataset is not big enough, then the model tends to overfit quickly, forgetting about the data that was used for pre-training and reducing the generalization ability of the final model. Hence, small starting learning rate and short training schedule are recommended.

  • If you would like to start training from pre-trained weights use --load-weights pararmeter.

    python train.py \
       --load-weights ${WORK_DIR}/snapshot.pth \
       --train-ann-files ${TRAIN_ANN_FILE} \
       --train-data-roots ${TRAIN_IMG_ROOT} \
       --val-ann-files ${VAL_ANN_FILE} \
       --val-data-roots ${VAL_IMG_ROOT} \
       --save-checkpoints-to ${WORK_DIR}/outputs

    Also you can use parameters such as --epochs, --batch-size, --gpu-num, --base-learning-rate, otherwise default values will be loaded from ${MODEL_TEMPLATE}.

  • If you would like to start fine-tuning from pre-trained weights use --resume-from parameter and value of --epochs have to exceed the value stored inside ${MODEL_TEMPLATE} file, otherwise training will be ended immediately. Here we add 5 additional epochs.

    export ADD_EPOCHS=5
    export EPOCHS_NUM=$((`cat ${MODEL_TEMPLATE} | grep epochs | tr -dc '0-9'` + ${ADD_EPOCHS}))
    
    python train.py \
       --resume-from ${WORK_DIR}/snapshot.pth \
       --train-ann-files ${TRAIN_ANN_FILE} \
       --train-data-roots ${TRAIN_IMG_ROOT} \
       --val-ann-files ${VAL_ANN_FILE} \
       --val-data-roots ${VAL_IMG_ROOT} \
       --save-checkpoints-to ${WORK_DIR}/outputs \
       --epochs ${EPOCHS_NUM}

6. Evaluation

Evaluation procedure allows us to get quality metrics values and complexity numbers such as number of parameters and FLOPs.

To compute MS-COCO metrics and save computed values to ${WORK_DIR}/metrics.yaml run:

python eval.py \
   --load-weights ${WORK_DIR}/outputs/latest.pth \
   --test-ann-files ${VAL_ANN_FILE} \
   --test-data-roots ${VAL_IMG_ROOT} \
   --save-metrics-to ${WORK_DIR}/metrics.yaml

You can also save images with predicted bounding boxes using --save-output-to parameter.

python eval.py \
   --load-weights ${WORK_DIR}/outputs/latest.pth \
   --test-ann-files ${VAL_ANN_FILE} \
   --test-data-roots ${VAL_IMG_ROOT} \
   --save-metrics-to ${WORK_DIR}/metrics.yaml \
   --save-output-to ${WORK_DIR}/output_images

If you have WiderFace dataset downloaded you also can specify --wider-dir parameter where WIDER_val.zip file is stored in order to compute official WiderFace metrics.

python eval.py \
   --load-weights ${WORK_DIR}/outputs/latest.pth \
   --test-ann-files ${VAL_ANN_FILE} \
   --test-data-roots ${VAL_IMG_ROOT} \
   --save-metrics-to ${WORK_DIR}/metrics.yaml \
   --wider-dir ${DATA_DIR}

7. Export PyTorch* model to the OpenVINO™ format

To convert PyTorch* model to the OpenVINO™ IR format run the export.py script:

python export.py \
   --load-weights ${WORK_DIR}/outputs/latest.pth \
   --save-model-to ${WORK_DIR}/export

This produces model model.xml and weights model.bin in single-precision floating-point format (FP32). The obtained model expects normalized image in planar BGR format.

For SSD networks an alternative OpenVINO™ representation is saved automatically to ${WORK_DIR}/export/alt_ssd_export folder. SSD model exported in such way will produce a bit different results (non-significant in most cases), but it also might be faster than the default one. As a rule SSD models in Open Model Zoo are exported using this option.

8. Validation of IR

Instead of passing snapshot.pth you need to pass path to model.bin (or model.xml).

python eval.py \
   --load-weights ${WORK_DIR}/export/model.bin \
   --test-ann-files ${VAL_ANN_FILE} \
   --test-data-roots ${VAL_IMG_ROOT} \
   --save-metrics-to ${WORK_DIR}/metrics.yaml