Update README.md

Author: geekysethi

README.md

@@ -1,43 +1,52 @@
-WHENet: Real-time Fine-Grained Estimation for Wide Range Head Pose
-===
-**Yijun Zhou and James Gregson - BMVC2020**
+# State of art the Head Pose Estimation in Tensorflow2 
 
+This repository includes:
+- ["WHENet: Real-time Fine-Grained Estimation for Wide Range Head Pose" (BMVC 2020).](https://www.bmvc2020-conference.com/assets/papers/0907.pdf) adapted from the [original source code](https://github.com/Ascend-Research/HeadPoseEstimation-WHENet).
 
-**Abstract:** We present an end-to-end head-pose estimation network designed to predict Euler
-angles through the full range head yaws from a single RGB image. Existing methods
-perform well for frontal views but few target head pose from all viewpoints. This has
-applications in autonomous driving and retail. Our network builds on multi-loss approaches
-with changes to loss functions and training strategies adapted to wide range
-estimation. Additionally, we extract ground truth labelings of anterior views from a
-current panoptic dataset for the first time. The resulting Wide Headpose Estimation Network
-(WHENet) is the first fine-grained modern method applicable to the full-range of
-head yaws (hence wide) yet also meets or beats state-of-the-art methods for frontal head
-pose estimation. Our network is compact and efficient for mobile devices and applications. [**ArXiv**](https://arxiv.org/abs/2005.10353)
 
-## Demo
-We provided two use case of the WHENet, image input and video input in this repo. Please make sure you installed all the requirments before running the demo code by `pip install -r requirements.txt`. Additionally, please download the [YOLOv3](https://drive.google.com/file/d/1wGrwu_5etcpuu_sLIXl9Nu0dwNc8YXIH/view?usp=sharing) model for head detection and put it under `yolo_v3/data`.
+- [RetinaFace: Single-stage Dense Face Localisation in the Wild](https://arxiv.org/abs/1905.00641) adapted from https://github.com/StanislasBertrand/RetinaFace-tf2.
 
-<img src=readme_imgs/video.gif height="220"/> <img src=readme_imgs/turn.JPG height="220"/> 
 
-## Image demo
-To run WHENet with image input, please put images and bbox.txt under one folder (E.g. Sample/) and just run `python demo.py`.
 
-Format of bbox.txt are showed below:
+
+
+<img src=images/output.png height="220"/> 
+
+
+
+## Install
+
+You can install this repository with pip (requires python>=3.6);
+
 ```
-image_name,x_min y_min x_max y_max
-mov_001_007585.jpeg,240 0 304 83
+pip install headpose_estimation
 ```
 
-## Video/Webcam demo
-We used [YOLO_v3](https://github.com/qqwweee/keras-yolo3) in the video demo to get the cropped head image. 
-In order to customize some of the functions we have put the yolo implementation and the pre-trained model in the repo.
-[Hollywood head](https://www.di.ens.fr/willow/research/headdetection/) and [Crowdhuman](https://www.crowdhuman.org/) are used to train the head detection YOLO model. 
-````
-demo_video.py [--video INPUT_VIDEO_PATH] [--snapshot WHENET_MODEL] [--display DISPLAY_OPTION] 
-              [--score YOLO_CONFIDENCE_THRESHOLD] [--iou IOU_THRESHOLD] [--gpu GPU#] [--output OUTPUT_VIDEO_PATH]
-````
-Please set `--video ''` for webcam input. 
+```bash
+pip install git+https://github.com/geekysethi/headpose_estimation
+```
+
+You can also install with the `setup.py`
+
+## Simple API with Face Detection
+To perform detection you can simple use the following lines:
+
+```python
+
+
+import cv2
+from headpose_estimation import Headpose
+
+if __name__ == "__main__":
+
+	headpose = Headpose()
+
+    img = cv2.imread("path_to_im.jpg")
+	detections,image = headpose.run(img)
+```
+
+This will return a list of dictionary which looks like this `[{'bbox': [xmin, ymin, xmax, ymax], 'yaw': yaw_value, 'pitch': pitch_value, 'roll': roll_value}`
+
 
 ## Dependncies
 * EfficientNet https://github.com/qubvel/efficientnet
-* Yolo_v3 https://github.com/qqwweee/keras-yolo3

images/output.png

@@ -1,12 +1,10 @@
 from setuptools import setup, find_packages
 import os
-
-# here = os.path.abspath(os.path.dirname(__file__))
-
-
-VERSION = '0.0.1'
+from pathlib import Path
+this_directory = Path(__file__).parent
+long_description = (this_directory / "README.md").read_text()
+VERSION = '0.0.2'
 DESCRIPTION = 'Head pose estimation module'
-LONG_DESCRIPTION = 'A package that allows to build simple streams of video, audio and camera data.'
 
 # Setting up
 setup(
@@ -16,18 +14,12 @@
     author_email="<ashish18024@iiitd.ac.in>",
     description=DESCRIPTION,
     long_description_content_type="text/markdown",
-    long_description=LONG_DESCRIPTION,
+    long_description=long_description,
     packages=find_packages(),
-    install_requires=['opencv-python', 'tensorflow-macos',],
-    keywords=['python', 'video', 'stream', 'video stream', 'camera stream', 'sockets'],
-    classifiers=[
-        "Development Status :: 1 - Planning",
-        "Intended Audience :: Developers",
-        "Programming Language :: Python :: 3",
-        "Operating System :: Unix",
-        "Operating System :: MacOS :: MacOS X",
-        "Operating System :: Microsoft :: Windows",
-    ],
+    install_requires=['', 'tensorflow-macos',],
+    keywords=['python', 'image', 'face detection', 'headpose estimation', 'machine learning', 'computer vision'],
+
+
     include_package_data=True,
 
     package_dir= {"cython":"headpose_estimation/face_detector/rcnn/cython/"}

setup.py

@@ -8,4 +8,10 @@
 	image_path = '/Users/ashish/Desktop/projects/HeadPoseEstimation-WHENet/Sample/random_internet_selfie.jpg'
 	img = cv2.imread(image_path)
 
-	headpose.run(img)
+	output,image = headpose.run(img)
+	print(output)
+
+	cv2.imwrite("output.png",image)
+	# cv2.imshow("image",image)
+	# cv2.waitKey()
+	# cv2.destroyAllWindows()