deep vision

deep vision is open-source software, created to help deep learning practitioners in applying computer vision techniques to videos and images, by doing most of the 'programming' part, and let them focus on the deep learning part.

Key Features

• Get images and videos from Camera, Files and Network.
• Display images and videos on desktop or web(work in progress).
• save resulting images and videos to file.
• Very easy to configure
• Very easy to add you deep learning model

Config file

Config Key	Default Value	Possible Values	Explanation
APP_NAME	'Deep Vision'	any string	The application title (desktop and web).
ABOUT_TITLE	"Deep Vision"	any string	The about box title (desktop and web).
ABOUT_MESSAGE	"Created By : FastAI Student"	any string	The about box contents (desktop and web).
APP_TYPE	'desktop'	desktop, web	Run the app as a desktop or a web application.
SHOW_MEDIA	True	True, False	Show images and videos on screen.
SAVE_MEDIA	True	True, False	Save images and videos to file.
SAVE_MEDIA_PATH	None	any string, None	Save images and videos to this folder, set to None to use application path.
SAVE_MEDIA_FILE	None	any string, None	Save images and videos to this file, set to None to use random file name.
SHOW_LOGS	False	True, False	Print applications logs to console.
CAMERA_ID	0	numbers	Target camera id, 0 for default camera
IMAGES_TYPES	['.jpg', '.jpeg', '.jpe', '.png', '*.bmp']	any image format supported by openCV	Open image dialog file types
VIDEOS_TYPES	['.mp4', '.avi']	any video format supported by openCV	Open video dialog file types
SCALE_IMAGES	True	True, False	Scale source image to fit container (desktop or web)

How to use

• install the dependencies

$ conda install pyqt
$ conda install opencv

• modify the config file to whatever you want
• open model.py file and follow the comments to add your model

from deep_vision.models_helpers.AbstractModel import AbstractModel
from deep_vision.models_helpers.imageutilities import put_text  # in future, this module will contain more helpers


class Model(AbstractModel):
    def __init__(self):
        # this code will be executed one time, when the user select a new image or video source.
        # initialize your model here
        # self.model = load_my_model()
        pass

    def predict(self, current_frame):
        # this code will be executed for every frame in the selected video, 
        # or only one time if the selected source is an image(an image is a single frame).
        # Get your model predictions for the current frame
        # preds = self.model.predict(image)
        # Then, draw boxes, write text, or do what ever you want to the current frame,
        # you can use the built-in model helpers, like put_text().
        put_text(current_frame, "preds")
        # return the frame, and see it displayed or saved
        return current_frame

• run the app