Matterport, a 3D camera maker, has just released a new deep learning algorthim under an open source license for pixel based object detection and segmentation.
CEO Matt Bell announced the news on Facebook November 1st.
He said “Matterport has just open-sourced our implementation of Mask R-CNN, a leading deep learning algorithm for doing pixel-wise object detection and segmentation in images. Compared to basic image detection algorithms, this provides much more precision and flexibility to do complex tasks.
“This is some amazing work by Waleed Abdulla, who implemented it from the rough information in the Facebook Research paper from a few months ago.
“You can find it here: https://github.com/matterport/Mask_RCNN
“If you’d like to help make more people aware of this great free open source tool, please upvote on hacker news and r/machinelearning. By working together we can finally bring the era of manual sheep-counting to an end.”
Matt Bell “Places to go to upvote (copy and paste the links, then look for the Mask R-CNN post):
Thank you Matt Bell!