BLOGS WEBSITE

Category: Machine Learning

ACVT has strong capabilities in the field of machine learning and is developing new techniques and applying them to a range of diverse problem spaces. For example, work is underway in the area of automatic feature extraction and classification from satellite and aerial imagery using a supervised machine learning system. Of key importance is the development of robust classifiers and an associated framework by which to apply them to geospatial data. ACVT is also working on entity extraction and resolution from unstructured data sources. This involves the development of unsupervised clustering techniques which are able to be applied on a massive scale.

We’re number one in VQA 2.0

A team led by Damien Teney (ACVT) and Peter Anderson (ACRV, ANU, and Microsoft) has just placed first in the VQA 2.0 challenge. Other members of the team include David Golub from Stanford, Po-Seng Huang, Lei Zhang and Xiaodong He from Microsoft, and Anton van den Hengel from ACVT. The leaderboard is here.

Posted in Machine Learning, News, Visual Question Answering | Tagged , , , , |

Comments Off on We’re number one in VQA 2.0

Medical machine learning on the Conversation

We just had a piece on medical machine learning published in the Conversation.

Posted in Machine Learning, Medical Imaging, News | Tagged , , |

Comments Off on Medical machine learning on the Conversation

Number 2 in ImageNet Scene Parsing Challenge 2016

We’ve had another great year in the ImageNet competition.  We came 2nd in the Scene Parsing challenge, which requires pixelwise segmentation of a large set of images into 150 classes of things and stuff.  The ImageNet Challenge is one of the most hotly contested challenges in Computer Vision, and is constantly updated to reflect the current […]

Posted in Events, Machine Learning, News, Segmentation | Tagged , , , , |

Comments Off on Number 2 in ImageNet Scene Parsing Challenge 2016

A new Machine Learning result in Quantum Physics

John Bastian and Anton van den Hengel are among the authors of a new paper just published in Nature Scientific Reports. The paper describes a Machine Learning-based approach to generating Bose-Einstein Condensates (BECs).  One of the difficulties in generating BECs is the need to cool Rubidium ions down to 10^-9 degrees Kelvin.  This is achieved […]

Posted in Machine Learning, News, Publication | Tagged |

Comments Off on A new Machine Learning result in Quantum Physics

We beat Google at ImageNet Detection

The ImageNet Object Detection results are out, and we did extremely well! We place 4th, behind Microsoft, Qualcomm, and Chinese University of Hong Kong, but ahead of Google, Intel, and Tencent. This is all the more impressive as we had a fraction of the computing resources of these competitors.

Posted in Machine Learning, News, Object Recognition | Tagged , , |

Comments Off on We beat Google at ImageNet Detection

Great ImageNet Detection Results

Last week was the deadline for the ImageNet Large Scale Visual Recognition Challenge (ILSVRC 2015) large-scale object detection task. This is the primary challenge for image-based object detection.  The challenge requires that you detect 200 classes of objects in a set of test images. For each image, algorithms must produce a set of annotations (ci,si,bi)of […]

Posted in Machine Learning, News, Object Recognition, Research | Tagged , , , |

Comments Off on Great ImageNet Detection Results

In deep learning end-to-end training of segmentation is best

A research team (Dr. Guosheng Lin, Prof. Chunhua Shen, Prof. Ian Reid, Prof. Anton van den Hengel) at the School of Computer Science, The University of Adelaide developed innovative “Deep Structured Learning” techniques that set up the new state-of-the-art semantic image segmentation record in the PASCAL VOC Challenge, which is organised by the University of Oxford.  The Adelaide team […]

Posted in Machine Learning, News, Research, Segmentation | Tagged |

Comments Off on In deep learning end-to-end training of segmentation is best

State of the art protein-protein interaction prediction

In another indication that the Machine Learning behind most Computer Vision Problems has more general applicability, we have just had a paper accepted which shows that the approach we developed for pedestrian detection achieves the world’s best performance in predicting protein-protein interactions. This result is based on the data set labelled ‘Physical Interaction Task in […]

Posted in Machine Learning, News, Research | Tagged |

Comments Off on State of the art protein-protein interaction prediction

World’s best Coco

The Microsoft COCO Captioning Challenge is designed to spur the development of algorithms producing image captions that are informative and accurate. There are 18 teams all together and our attributes based image captioning framework currently achieves the best result on 3 evaluation metrics (BLEU-1,2,3) out of 7. We also achieve the top-5 ranking on the […]

Posted in Image Captioning, Machine Learning, News, Research | Tagged , |

Comments Off on World’s best Coco

ACVT semantic image segmentation technique tops in the PASCAL VOC Challenge

Researchers at ACVT have developed new “Deep Structured Learning” techniques that set up the new state-of-the-art semantic image segmentation record in the PASCAL VOC Challenge, which is organised by Oxford University. Semantic image segmentation is one of the tasks and probably the most challenging one, which is to label each pixel in images. Deep Learning is the […]

Posted in Machine Learning, News | Tagged , |

Comments Off on ACVT semantic image segmentation technique tops in the PASCAL VOC Challenge