ArtificialVision 0.1.4

Artificial Vision Library

ArtificialVision

❗️ The package is still under development and has not been released yet ❗️

If the version exceeds 1.0.0, this message will be removed, and the package will become available.

Installation

pip install artificialvision

What is ArtificialVision?

ArtificialVision is the package for makes it easy to get the outcomes of various Machine Learning & Computer Vision technologies. This package's aims are improving the quality and increasing the productivity by using it for convenience in various research and development experiments.

In this Version, just inference & getting the various results are supported. Support for training and fine-tuning will be added in the future.

Contributing to ArtificialVision (Not yet)

All the contributions are welcome !

Check the ContributeGuide.md for more information.

TODO

Primary Methods

• Classification
• Object Detection
• Segmentation
• Pose Estimation
• feature Extraction

Secondary Methods

• Matching
• Tracking
• Generation
• Restoration
• Super-Resolution

Contributors

Methods Tutorial

Image Classification

example

from artificialvision import ImgClassification
import cv2 

# Read the image
img = cv2.imread('PATH of Image file')

# Get the classification result
ImgClassification.get_result(img)

Currently, only models pretrained on ImageNet are available.

Object Detection

example

from artificialvision import ObjDetection
import cv2

''' Image '''
# Read the image
img = cv2.imread('PATH of Image file')

# Get the detection result with the bounding box
ObjDetection.get_result(img)

# Get the bounding box only
ObjDetection.get_result_with_box(img)

''' Video '''
# Read the video
video = cv2.VideoCapture('PATH of Video file', type=1)

# Get the detection result with the bounding box
ObjDetection.get_result(video)

# Get the bounding box only
ObjDetection.get_result_with_box(video)

hyperparameters

• type : int, default is 0
  • 0 : Image
  • 1 : Video

Currently, only image and video matching are supported.

Segmentation

example

from artificialvision import Segmentation
import cv2

''' Image '''
# Read the image
img = cv2.imread('PATH of Image file')

# Get the segmentation result
Segmentation.get_result(img)

# Get only the segment map
Segmentation.get_segment_map(img)
''' Video '''
# Read the video
video = cv2.VideoCapture('PATH of Video file', type=1)

# Get the segmentation result
Segmentation.get_result(video)

# Get only the segment map
Segmentation.get_segment_map(video)
''' Webcam (real-time) '''
# start the webcam(recording)
# if finished, press 'q' to stop & get the result
Segmentation.get_result(type=2)

hyperparameters

• type : int, default is 0

  • 0 : Image
  • 1 : Video
  • 2 : Webcam (real-time)

• category : int, default is 0

  • segmentation category
  • 0 : Semantic Segmentation
  • 1 : Instance Segmentation
  • 2 : Panoptic Segmentation

• detail : int, default is 0

  • segmentation detail
  • 0 : Segmentation Result (Overlayed Image)
  • 1 : Segmentation Map

• get_poligon : bool, default is False

  • If True, get the poligon points of the segmentation result. (Only for the instance segmentation)

Currently, only image and video matching are supported.

Image Matching

example

from artificialvision import ImgMatching
import cv2 

''' Image '''
# Read the images
img1 = cv2.imread('PATH of Image1 file')
img2 = cv2.imread('PATH of Image2 file')

# Get the matching score
ImgMatching.get_matching_score(img1, img2)

# Get the matching result
ImgMatching.get_matching_result(img1, img2)


''' Video '''
# Read the videos
video1 = cv2.VideoCapture('PATH of Video1 file')
video2 = cv2.VideoCapture('PATH of Video2 file')

# Get the matching score
ImgMatching.get_matching_score(video1, video2, type=1)

# Get the matching result
ImgMatching.get_matching_result(video1, video2, type=1)

''' Mixed '''
# Read the images for matchin
img_list = [img1, img2, img3, ...]

# Get the matching score
ImgMatching.get_matching_score(img_list, video1, type=2)

# Get the matching result
ImgMatching.get_matching_result(img_list, video1, type=2)

''' Webcam (real-time) '''
# start the webcam(recording)
# if finished, press 'q' to stop & get the result
ImgMatching.get_matching_result(img_list, type=3)

hyperparameters

• type : int, default is 0
  • 0 : Image
  • 1 : Video
  • 2 : Mixed
  • 3 : Webcam (real-time)
• threshold : float, default is 0.5
  • The threshold for the matching score. If the matching score is less than the threshold, it is considered as a matching result. Range is 0.0 ~ 1.0. Recommended is +-0.1 from the default value.

Currently, only image and video matching are supported.

Format

Inference Data Format

Inference data format	Type in python	Usage Example
Path of the data	str	'/Path/to/data/file.extension'
List	list
Numpy Array	numpy.ndarray
Pytorch Tensor	torch.Tensor
Tensorflow Tensor	tensorflow.python.framework.ops.EagerTensor

Inference Model Format

Inference model format	export.py --include	Model
PyTorch	-	model.pt
TorchScript	torchscript	model.torchscript
ONNX	onnx	model.onnx
OpenVINO	openvino	model_openvino_model/
TensorRT	engine	model.engine
CoreML	coreml	model.mlmodel
TensorFlow SavedModel	saved_model	model_saved_model/
TensorFlow GraphDef	pb	model.pb
TensorFlow Lite	tflite	model.tflite
TensorFlow Edge TPU	edgetpu	model_edgetpu.tflite
TensorFlow.js	tfjs	model_web_model/
PaddlePaddle	paddle	model_paddle_model/

---

If you want more information, check the Official Docs(Not yet)