Hristo Bojinov

Ramkumar B, R. S. Hegde, Rob Laber et al.

The recently proposed open-source KAZE image feature detection and description algorithm offers unprecedented performance in comparison to conventional ones like SIFT and SURF as it relies on nonlinear scale spaces instead of Gaussian linear scale spaces. The improved performance, however, comes with a significant computational cost limiting its use for many applications. We report a GPGPU implementation of the KAZE algorithm without resorting to binary descriptors for gaining speedup. For a 1920 by 1200 sized image our Compute Unified Device Architecture (CUDA) C based GPU version took around 300 milliseconds on a NVIDIA GeForce GTX Titan X (Maxwell Architecture-GM200) card in comparison to nearly 2400 milliseconds for a multithreaded CPU version (16 threaded Intel(R) Xeon(R) CPU E5-2650 processsor). The CUDA based parallel implementation is described in detail with fine-grained comparison between the GPU and CPU implementations. By achieving nearly 8 fold speedup without performance degradation our work expands the applicability of the KAZE algorithm. Additionally, the strategies described here can prove useful for the GPU implementation of other nonlinear scale space based methods.

Manik Goyal, Param Rajpura, Hristo Bojinov et al.

Although Deep Convolutional Neural Networks trained with strong pixel-level annotations have significantly pushed the performance in semantic segmentation, annotation efforts required for the creation of training data remains a roadblock for further improvements. We show that augmentation of the weakly annotated training dataset with synthetic images minimizes both the annotation efforts and also the cost of capturing images with sufficient variety. Evaluation on the PASCAL 2012 validation dataset shows an increase in mean IOU from 52.80% to 55.47% by adding just 100 synthetic images per object class. Our approach is thus a promising solution to the problems of annotation and dataset collection.

Param S. Rajpura, Hristo Bojinov, Ravi S. Hegde

The need for large annotated image datasets for training Convolutional Neural Networks (CNNs) has been a significant impediment for their adoption in computer vision applications. We show that with transfer learning an effective object detector can be trained almost entirely on synthetically rendered datasets. We apply this strategy for detecting pack- aged food products clustered in refrigerator scenes. Our CNN trained only with 4000 synthetic images achieves mean average precision (mAP) of 24 on a test set with 55 distinct products as objects of interest and 17 distractor objects. A further increase of 12% in the mAP is obtained by adding only 400 real images to these 4000 synthetic images in the training set. A high degree of photorealism in the synthetic images was not essential in achieving this performance. We analyze factors like training data set size and 3D model dictionary size for their influence on detection performance. Additionally, training strategies like fine-tuning with selected layers and early stopping which affect transfer learning from synthetic scenes to real scenes are explored. Training CNNs with synthetic datasets is a novel application of high-performance computing and a promising approach for object detection applications in domains where there is a dearth of large annotated image data.

Hristo Bojinov, Yan Michalevsky, Gabi Nakibly et al.

We demonstrate how the multitude of sensors on a smartphone can be used to construct a reliable hardware fingerprint of the phone. Such a fingerprint can be used to de-anonymize mobile devices as they connect to web sites, and as a second factor in identifying legitimate users to a remote server. We present two implementations: one based on analyzing the frequency response of the speakerphone-microphone system, and another based on analyzing device-specific accelerometer calibration errors. Our accelerometer-based fingerprint is especially interesting because the accelerometer is accessible via JavaScript running in a mobile web browser without requesting any permissions or notifying the user. We present the results of the most extensive sensor fingerprinting experiment done to date, which measured sensor properties from over 10,000 mobile devices. We show that the entropy from sensor fingerprinting is sufficient to uniquely identify a device among thousands of devices, with low probability of collision.