#Gif machine learning code

Import numpy as np import os import coco import model as modellib import glob import imageio import cv2 # Root directory to project ROOT_DIR = os.

#Gif machine learning code

Something like the following code lets us extract just a transparent cutout of a person from each of our video’s images: In order to do something with the mask, you’ll need to use it as a channel to copy or paste another image into it. Each mask is 0, or black where there is no detected object, and 255 or white, where there is a detected object. In general, the masks are unsigned 8 bit integers, in the shape of your input image. Once you get the repo set up locally, I recommend running through the demo notebook, and exploring how the images and detection masks work. The Matterport Mask R-CNN comes with a few great Jupyter notebooks that help explore how it works. With that in order, let’s jump into detecting and manipulating our images with Mask R-CNN. We can set this lower if we want to slow down or video, or higher if we want to speed it up. The -r parameter specifies the frames per second we want to use to build our output video. $ ffmpeg -r 60 -f image2 -i %05d.jpg OUTPUT.mp4 You’ll need to keep track of your input video’s frames per second in order to keep things synced up after conversion. Depending on your video input source, it may be between 24 and 60 frames per second. With the following command, we get a sequence of images out from an input video. Working with Video in PythonĪlthough there are a few great options for working with video in Python, I prefer to convert my videos directly into image sequences and back again using ffmpeg. So for that, we’ll use this video I recorded of myself walking around in our back yard. We’ll be able to see when the human gets lost, and how noisey the detection outline model really is. That way we can isolate and see how well our model can mask for a single individual. Rather than begin by trying to detect and track multiple people for a gif creator, first start with an input video with just one human. To begin with my automatic gif generator, I’ve opted to do the most stupid thing first, a pattern I’ve found works well in creative coding projects. It works with Python 3, it’s got great example code, and it’s probably the easiest version to install.

There’s been a few versions of the Mask R-CNN paper implemented, but lately I’ve been playing with the MatterPort version. Today, we’ll use a new deep learning model, called Mask R-CNN to extract object gifs automatically from video. Specifically, we used a deep learning model called “ YOLO”.

In previous posts, we used deep learning to automatically detect and record birds from a webcam.