10 (more) Deep Learning projects based on Apache MXNet
In a previous article, I listed 10 cool Deep Learning projects based on Apache MXNet. Well, here are 10 more, a nice mix of model implementations and applications.
If you have an MXNet project that I haven’t listed to far, please get in touch!
Model implementations
#1 — DenseNet
This is an implementation of the DenseNet-BC architecture as described in the Densely Connected Convolutional Networks, by by Gao Huang, Zhuang Liu, Kilian Q. Weinberger and Laurens van der Maaten.
This architecture contains shorter connections between layers close to the input and those close to the output. They help models train more efficiently and predict more accurately.
#2 — Binary Neural Networks
This project implement Binary Neural Networks, as described in BMXNet: An Open-Source Binary Neural Network Implementation Based on MXNet by Haojin Yang, Martin Fritzsche, Christian Bartz and Christoph Meinel.
These networks use weights are binary values! At the cost of minimal accuracy loss, these networks are both much smaller and much faster than their floating-point counterparts.
#3 — Mask R-CNN (image segmentation)
This is an implementation of the Mask R-CNN architecture, based on the self-titled paper by Kaiming He, Georgia Gkioxari, Piotr Dollár and Ross Girshick
This architecture is an evolution of Fast R-CNN and does a very good job at object segmentation. If case you didn’t know, TuSimple build autonomous driving systems :)
#4 — YOLO9000 (object detection)
This project performs object detection based on the YOLO9000: Better, Faster, Stronger research paper by Joseph Redmon and Ali Farhadi.
At 40 frames per second, YOLOv2 gets 78.6 mean average precision, “outperforming state-of-the-art methods like Faster RCNN with ResNet and SSD while still running significantly faster”.
#5— STN-OCR (text detection and text recognition)
This project implements the model described in STN-OCR: A single Neural Network for Text Detection and Text Recognition, by Christian Bartz, Haojin Yang and Christoph Meinel.
Applications
#6— Head pose estimation
This model is a simple CNN that does a good job at detecting head poses.
#7 — Realtime multi-person pose estimation
This project implements an MXNet version of the model described in Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields by Zhe Cao, Tomas Simon, Shih-En Wei and Yaser Sheikh.
This is AMAZING.
#8 — Sentiment analysis
This tutorial shows you how to build a sentiment analysis model , based on Convolutional Neural Networks for Sentence Classification by Yoon Kim. The author provides clear notebooks using both Keras and MXNet. Very nicely done!
#9 & #10 —Image detection on mobile
Let’s not forget mobile application developers. These twin projects show you how to use an MXNet model in Android and iOS apps.
That’s it for today. Thanks for reading.
This article was written while overdosing on Rainbow albums. A Light in the Black… you can’t top this.
