Arun Mallya

I am a third fourth year PhD student in the vision lab at the University of Illinois at Urbana-Champaign exploring computer vision under the guidance of Prof. Svetlana Lazebnik.

I previously completed my Master's in CS at UIUC, and my Bachelor's in CSE at IIT KGP.

email / github / LinkedIn / webpage



A M
Research

PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning
Arun Mallya, Svetlana Lazebnik
[arxiv preprint]

Recurrent Models for Situation Recognition
Arun Mallya, Svetlana Lazebnik
International Conference on Computer Vision (ICCV), 2017
[arxiv preprint] [poster]

Phrase Localization and Visual Relationship Detection with Comprehensive Linguistic Cues
Bryan A. Plummer, Arun Mallya, Christopher M. Cervantes, Julia Hockenmaier, Svetlana Lazebnik
International Conference on Computer Vision (ICCV), 2017
[arxiv preprint]

A torch implementation of the MLP VQA system, along with data from
Revisiting Visual Question Answering Baselines, ECCV'16, can be found here.

Learning Models for Actions and Person-Object Interactions with Transfer to Question Answering
Arun Mallya, Svetlana Lazebnik
European Conference on Computer Vision (ECCV), 2016
[arxiv preprint] [caffemodels]

Solving Visual Madlibs with Multiple Cues
Tatiana Tommasi, Arun Mallya, Bryan Plummer, Svetlana Lazebnik, Alex Berg, Tamara Berg
British Machine Vision Conference (BMVC), 2016
[arxiv preprint]
International Journal of Computer Vision (IJCV), 2016, submitted
[arxiv preprint]

Trained models: [caffemodels]

  1. The Fusion model from the ECCV'16 paper above, trained on the HICO action dataset to predict 600 human-object interactions.
  2. The Fusion model from the ECCV'16 paper above, trained on the MPII action dataset to predict 393 human actions.
  3. A VGG-16 model to predict 302 person attributes selected from the Flickr30k Entities dataset.

Learning Informative Edge Maps for Indoor Scene Layout Prediction
Arun Mallya, Svetlana Lazebnik
International Conference on Computer Vision (ICCV), 2015
[pdf] [hedau+ dataset] [hedau+ processed data]
[hedau+ LMDBs (14GB)] [trained model + demo]

Part Localization using Multi-Proposal Consensus for Fine-Grained Categorization
Kevin J. Shih, Arun Mallya, Saurabh Singh, Derek Hoiem
British Machine Vision Conference (BMVC), 2015
[arxiv preprint] [poster] [webpage]

Unsupervised Deep Network Pretraining via Human Design
Ming-Yu Liu, Arun Mallya, Oncel C. Tuzel, Xi Chen
Winter Conference on Applications of Computer Vision (WACV), 2016
Work performed as a summer intern at MERL Cambridge
[arxiv preprint]


Workshop Submissions
  1. High-level Cues for Predicting Motivations (poster) (slides)
    Arun Mallya, Svetlana Lazebnik
    Scene Understanding Workshop, CVPR 2017

  2. Visual Relationship Detection with Multiple Cues (poster)
    Arun Mallya, Bryan A. Plummer, Svetlana Lazebnik
    Language and Vision Workshop, CVPR 2017

Writeups/Notes
Hosted on github. Edit requests/additions/corrections are welcome.
  1. A Backpropagation Refresher
  2. An Illustrated Explanation of the LSTM Forward-Backward Pass
  3. Introduction to RNNs
  4. Introduction to RNNs - II
  5. Jupyter notebook to find Receptive Field Size and Effective Stride (supports dilated convs)
  6. Visualization of neuron connections and receptive field of a CNN (including dilation)!

(imitation is the sincerest form of flattery)