# Batch Face for Modern Research
## 🚧Documentation under construction, check tests folder for more details. 🚧
This repo provides the out-of-box face detection and face alignment with batch input support and enables real-time application on CPU.
## Features
1. Batch input support for faster data processing.
2. Smart API.
3. Ultrafast with inference runtime acceleration.
4. Automatically download pre-trained weights.
5. Minimal dependencies.
6. Unleash the power of GPU for batch processing.
### Requirements
- Linux, Windows or macOS
- Python 3.5+ (it may work with other versions too)
- opencv-python
- PyTorch (>=1.0)
- ONNX (optional)
While not required, for optimal performance it is highly recommended to run the code using a CUDA enabled GPU.
## Install
The easiest way to install it is using pip:
```bash
pip install git+https://github.com/elliottzheng/batch-face.git@master
```
No extra setup needs, most of the pretrained weights will be downloaded automatically.
If you have trouble install from source, you can try install from PyPI:
```bash
pip install batch-face
```
the PyPI version is not guaranteed to be the latest version, but we will try to keep it up to date.
## Usage
You can clone the repo and run tests like this
```
python -m tests.camera
```
### Face Detection
##### Detect face and five landmarks on single image
```python
import cv2
from batch_face import RetinaFace
detector = RetinaFace(gpu_id=0)
img = cv2.imread("examples/obama.jpg")
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
max_size = 1080 # if the image's max size is larger than 1080, it will be resized to 1080, -1 means no resize
resize = 1 # resize the image to speed up detection, default is 1, no resize
threshold = 0.95 # confidence threshold
# now we recommand to specify return_dict=True to get the result in a more readable way
faces = detector(img, threshold=threshold, resize=resize, max_size=max_size, return_dict=True)
face = faces[0]
box = face['box']
kps = face['kps']
score = face['score']
# the old way to get the result
faces = detector(img, threshold=threshold, resize=resize, max_size=max_size)
box, kps, score = faces[0]
```
##### Running on CPU/GPU
In order to specify the device (GPU or CPU) on which the code will run one can explicitly pass the device id.
```python
from batch_face import RetinaFace
# 0 means using GPU with id 0 for inference
# default -1: means using cpu for inference
fp16 = True # use fp16 to speed up detection and save GPU memory
detector = RetinaFace(gpu_id=0, fp16=True)
```
| | GPU(GTX 1080TI,batch size=1) | GPU(GTX 1080TI，batch size=750) | CPU(Intel(R) Core(TM) i7-7800X CPU @ 3.50GHz) |
| ---- | ---------------------------- | ------------------------------- | --------------------------------------------- |
| FPS | 44.02405810720893 | 96.64058005582535 | 15.452635835550483 |
| SPF | 0.022714852809906007 | 0.010347620010375976 | 0.0647138786315918 |
##### Batch input for faster detection
**Detector with CUDA process batch input faster than the same amount of single input.**
```python
import cv2
from batch_face import RetinaFace
detector = RetinaFace()
img= cv2.imread('examples/obama.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
max_size = 1080 # if the image's max size is larger than 1080, it will be resized to 1080, -1 means no resize
resize = 1 # resize the image to speed up detection, default is 1, no resize
resize_device = 'cpu' # resize on cpu or gpu, default is gpu
threshold = 0.95 # confidence threshold for detection
batch_size = 100 # batch size for detection, the larger the faster but more memory consuming, default is -1, which means batch_size = number of input images
batch_images = [img,img] # pseudo batch input
all_faces = detector(batch_images, threshold=threshold, resize=resize, max_size=max_size, batch_size=batch_size)
faces = all_faces[0] # the first input image's detection result
box, kps, score = faces[0] # the first face's detection result
```
Note: All the input images must of the same size, for input images with different size, please use `detector.pseudo_batch_detect`.
### Face Alignment
##### face alignment on single image
```python
from batch_face import drawLandmark_multiple, LandmarkPredictor, RetinaFace
predictor = LandmarkPredictor(0)
detector = RetinaFace(0)
imgname = "examples/obama.jpg"
img = cv2.imread(imgname)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
faces = detector(img)
if len(faces) == 0:
print("NO face is detected!")
exit(-1)
# the first input for the predictor is a list of face boxes. [[x1,y1,x2,y2]]
results = predictor(faces, img, from_fd=True) # from_fd=True to convert results from our detection results to simple boxes
for face, landmarks in zip(faces, results):
img = drawLandmark_multiple(img, face[0], landmarks)
```
### Head Pose Estimation
##### Head pose estimation on video
```python
from batch_face import RetinaFace, SixDRep, draw_landmarks, load_frames_rgb, Timer
vis = True
gpu_id = 0
batch_size = 100
threshold = 0.95
detector = RetinaFace(gpu_id=gpu_id)
head_pose_estimator = SixDRep(gpu_id=gpu_id)
video_file = 'examples/ross.mp4'
frames = load_frames_rgb(video_file) # simple wrapper to load video frames with opencv and convert to RGB, 0~255, UInt8, HWC
print(f'Loaded {len(frames)} frames')
print('image size:', frames[0].shape)
# it might take longer time to detect since is first time to run the model
all_faces = detector(frames, batch_size=batch_size, return_dict=True, threshold=threshold, resize=0.5)
head_poses = head_pose_estimator(all_faces, frames, batch_size=batch_size, update_dict=True, input_face_type='dict')
# the head pose will be updated in the all_faces dict
out_frames = []
for faces, frame in zip(all_faces, frames):
for face in faces:
head_pose_estimator.plot_pose_cube(frame, face['box'], **face['head_pose'])
out_frames.append(frame)
if vis:
import imageio
out_file = 'examples/head_pose.mp4'
imageio.mimsave(out_file, out_frames, fps=8)
```
check out the result video [here](./examples/head_pose.mp4)
you can run the script `python -m tests.video_head_pose` to see the result.
## References
- Face Detection Network and pretrained model are from [biubug6/Pytorch_Retinaface](https://github.com/biubug6/Pytorch_Retinaface)
- Face Alignment Network and pretrained model are from [cunjian/pytorch_face_landmark](https://github.com/cunjian/pytorch_face_landmark)
- Face Reconstruction Network and pretrained model are from [cleardusk/3DDFA](https://github.com/cleardusk/3DDFA)
- Head Pose Estimation Network and pretrained model are from [jahongir7174/SixDRepNet](https://github.com/jahongir7174/SixDRepNet)
Raw data
{
"_id": null,
"home_page": "https://github.com/elliottzheng/batch-face",
"name": "batch-face",
"maintainer": null,
"docs_url": null,
"requires_python": null,
"maintainer_email": null,
"keywords": "face-detection pytorch RetinaFace face-alignment",
"author": "Elliott Zheng",
"author_email": "admin@hypercube.top",
"download_url": null,
"platform": null,
"description": "# Batch Face for Modern Research\n\n## \ud83d\udea7Documentation under construction, check tests folder for more details. \ud83d\udea7\n\nThis repo provides the out-of-box face detection and face alignment with batch input support and enables real-time application on CPU.\n\n## Features\n1. Batch input support for faster data processing.\n2. Smart API.\n3. Ultrafast with inference runtime acceleration.\n4. Automatically download pre-trained weights.\n5. Minimal dependencies.\n6. Unleash the power of GPU for batch processing.\n\n### Requirements\n\n- Linux, Windows or macOS\n- Python 3.5+ (it may work with other versions too)\n- opencv-python\n- PyTorch (>=1.0) \n- ONNX (optional)\n\nWhile not required, for optimal performance it is highly recommended to run the code using a CUDA enabled GPU.\n\n## Install\n\nThe easiest way to install it is using pip:\n\n```bash\npip install git+https://github.com/elliottzheng/batch-face.git@master\n```\nNo extra setup needs, most of the pretrained weights will be downloaded automatically.\n\nIf you have trouble install from source, you can try install from PyPI:\n\n```bash\npip install batch-face\n```\nthe PyPI version is not guaranteed to be the latest version, but we will try to keep it up to date.\n\n## Usage\nYou can clone the repo and run tests like this\n```\npython -m tests.camera\n```\n### Face Detection\n\n##### Detect face and five landmarks on single image\n```python\nimport cv2\nfrom batch_face import RetinaFace\n\ndetector = RetinaFace(gpu_id=0)\nimg = cv2.imread(\"examples/obama.jpg\")\nimg = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n\nmax_size = 1080 # if the image's max size is larger than 1080, it will be resized to 1080, -1 means no resize\nresize = 1 # resize the image to speed up detection, default is 1, no resize\nthreshold = 0.95 # confidence threshold\n\n# now we recommand to specify return_dict=True to get the result in a more readable way\nfaces = detector(img, threshold=threshold, resize=resize, max_size=max_size, return_dict=True)\nface = faces[0]\nbox = face['box']\nkps = face['kps']\nscore = face['score']\n\n# the old way to get the result\nfaces = detector(img, threshold=threshold, resize=resize, max_size=max_size)\nbox, kps, score = faces[0]\n\n```\n##### Running on CPU/GPU\n\nIn order to specify the device (GPU or CPU) on which the code will run one can explicitly pass the device id.\n```python\nfrom batch_face import RetinaFace\n# 0 means using GPU with id 0 for inference\n# default -1: means using cpu for inference\nfp16 = True # use fp16 to speed up detection and save GPU memory\n\ndetector = RetinaFace(gpu_id=0, fp16=True)\n```\n| | GPU(GTX 1080TI,batch size=1) | GPU(GTX 1080TI\uff0cbatch size=750) | CPU(Intel(R) Core(TM) i7-7800X CPU @ 3.50GHz) |\n| ---- | ---------------------------- | ------------------------------- | --------------------------------------------- |\n| FPS | 44.02405810720893 | 96.64058005582535 | 15.452635835550483 |\n| SPF | 0.022714852809906007 | 0.010347620010375976 | 0.0647138786315918 |\n\n\n##### Batch input for faster detection\n\n**Detector with CUDA process batch input faster than the same amount of single input.** \n\n```python\nimport cv2\nfrom batch_face import RetinaFace\n\ndetector = RetinaFace()\nimg= cv2.imread('examples/obama.jpg')\nimg = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n\nmax_size = 1080 # if the image's max size is larger than 1080, it will be resized to 1080, -1 means no resize\nresize = 1 # resize the image to speed up detection, default is 1, no resize\nresize_device = 'cpu' # resize on cpu or gpu, default is gpu\nthreshold = 0.95 # confidence threshold for detection\nbatch_size = 100 # batch size for detection, the larger the faster but more memory consuming, default is -1, which means batch_size = number of input images\nbatch_images = [img,img] # pseudo batch input\n\nall_faces = detector(batch_images, threshold=threshold, resize=resize, max_size=max_size, batch_size=batch_size) \nfaces = all_faces[0] # the first input image's detection result\nbox, kps, score = faces[0] # the first face's detection result\n```\n\nNote: All the input images must of the same size, for input images with different size, please use `detector.pseudo_batch_detect`.\n\n\n\n### Face Alignment\n##### face alignment on single image\n\n```python \nfrom batch_face import drawLandmark_multiple, LandmarkPredictor, RetinaFace\n\npredictor = LandmarkPredictor(0)\ndetector = RetinaFace(0)\n\nimgname = \"examples/obama.jpg\"\nimg = cv2.imread(imgname)\nimg = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n\nfaces = detector(img)\n\nif len(faces) == 0:\n print(\"NO face is detected!\")\n exit(-1)\n\n# the first input for the predictor is a list of face boxes. [[x1,y1,x2,y2]]\nresults = predictor(faces, img, from_fd=True) # from_fd=True to convert results from our detection results to simple boxes\n\nfor face, landmarks in zip(faces, results):\n img = drawLandmark_multiple(img, face[0], landmarks)\n```\n### Head Pose Estimation\n##### Head pose estimation on video\n```python\nfrom batch_face import RetinaFace, SixDRep, draw_landmarks, load_frames_rgb, Timer\n\nvis = True\ngpu_id = 0\nbatch_size = 100\nthreshold = 0.95\ndetector = RetinaFace(gpu_id=gpu_id)\nhead_pose_estimator = SixDRep(gpu_id=gpu_id)\nvideo_file = 'examples/ross.mp4'\nframes = load_frames_rgb(video_file) # simple wrapper to load video frames with opencv and convert to RGB, 0~255, UInt8, HWC\nprint(f'Loaded {len(frames)} frames')\nprint('image size:', frames[0].shape)\n# it might take longer time to detect since is first time to run the model\nall_faces = detector(frames, batch_size=batch_size, return_dict=True, threshold=threshold, resize=0.5)\nhead_poses = head_pose_estimator(all_faces, frames, batch_size=batch_size, update_dict=True, input_face_type='dict')\n# the head pose will be updated in the all_faces dict\nout_frames = []\nfor faces, frame in zip(all_faces, frames):\n for face in faces:\n head_pose_estimator.plot_pose_cube(frame, face['box'], **face['head_pose'])\n out_frames.append(frame)\n\nif vis:\n import imageio\n out_file = 'examples/head_pose.mp4'\n imageio.mimsave(out_file, out_frames, fps=8)\n```\ncheck out the result video [here](./examples/head_pose.mp4)\nyou can run the script `python -m tests.video_head_pose` to see the result.\n\n\n## References\n\n- Face Detection Network and pretrained model are from [biubug6/Pytorch_Retinaface](https://github.com/biubug6/Pytorch_Retinaface)\n- Face Alignment Network and pretrained model are from [cunjian/pytorch_face_landmark](https://github.com/cunjian/pytorch_face_landmark)\n- Face Reconstruction Network and pretrained model are from [cleardusk/3DDFA](https://github.com/cleardusk/3DDFA)\n- Head Pose Estimation Network and pretrained model are from [jahongir7174/SixDRepNet](https://github.com/jahongir7174/SixDRepNet)\n",
"bugtrack_url": null,
"license": "MIT",
"summary": "Batch Face Preprocessing for Modern Research",
"version": "1.5.0",
"project_urls": {
"Documentation": "https://github.com/elliottzheng/batch-face",
"Homepage": "https://github.com/elliottzheng/batch-face",
"Source": "https://github.com/elliottzheng/batch-face",
"Tracker": "https://github.com/elliottzheng/batch-face/issues"
},
"split_keywords": [
"face-detection",
"pytorch",
"retinaface",
"face-alignment"
],
"urls": [
{
"comment_text": "",
"digests": {
"blake2b_256": "53ef18071a0584ed1133766e45da69f2aedc3ca2377105e4de84f5d46bfc8101",
"md5": "0135dd6635592a5b2fb255a15dbfd04c",
"sha256": "ae13eea8e2f58731f28ee0b9f4c14b3fd6c959864a0e2c249f23be683c432e76"
},
"downloads": -1,
"filename": "batch_face-1.5.0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "0135dd6635592a5b2fb255a15dbfd04c",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": null,
"size": 30623534,
"upload_time": "2024-08-01T06:06:32",
"upload_time_iso_8601": "2024-08-01T06:06:32.731199Z",
"url": "https://files.pythonhosted.org/packages/53/ef/18071a0584ed1133766e45da69f2aedc3ca2377105e4de84f5d46bfc8101/batch_face-1.5.0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2024-08-01 06:06:32",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "elliottzheng",
"github_project": "batch-face",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"requirements": [
{
"name": "numpy",
"specs": []
},
{
"name": "torch",
"specs": []
},
{
"name": "torchvision",
"specs": []
},
{
"name": "opencv_transforms",
"specs": []
}
],
"lcname": "batch-face"
}
JSON
