Dolphin Recognition with Adaptive Hybrid Saliency Detection for Deep Learning Based on DenseNet Recognition

Dolphin identification is important for wildlife conservation. Since identifying dolphins from thousands of images manually takes tremendous time, it is important to develop an automatic dolphin identification algorithm. In this paper, a high accurate deep learning based dolphin identification algorithm is proposed. We presented an advanced approach, called hybrid saliency method, for feature extraction and efficiently integrate several well-known techniques to make dolphins distinguishable. With the proposed techniques, we can avoid the background part (e.g. the sea water) to affect the identification results, which is usually a problem of most convolutional neural network based methods. Simulations show that the proposed algorithm can well identify a dolphin in most cases and it can achieve the accuracy rate of 85% even if there are 40 dolphins to be distinguished.

[1]  Mahmoud R. El-Sakka,et al.  Improving Right Whale recognition by fine-tuning alignment and using wide localization network , 2017, 2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE).

[2]  B. White,et al.  DIFFERENCES IN THE EXTERNAL MORPHOLOGY OF TWO SYMPATRIC SPECIES OF BOTTLENOSE DOLPHINS (GENUS TURSIOPS) IN THE WATERS OF CHINA , 2000 .

[3]  Andrei Mihai,et al.  Right whale recognition using convolutional neural networks , 2016, ArXiv.

[4]  Babak Nadjar Araabi,et al.  Computer-assisted photo-identié cation of individual marine vertebrates: a multi-species system , 2003 .

[5]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[6]  V. Akhil.  SALIENT REGION DETECTION via HIGH-DIMENSIONAL COLOR TRANSFORM AND LOCAL SPATIAL SUPPORT , 2018 .

[7]  Kilian Q. Weinberger,et al.  Densely Connected Convolutional Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Zhuowen Tu,et al.  Deeply Supervised Salient Object Detection with Short Connections , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[9]  Zhiming Luo,et al.  Non-local Deep Features for Salient Object Detection , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  Charles V. Stewart,et al.  Integral Curvature Representation and Matching Algorithms for Identification of Dolphins and Whales , 2017, 2017 IEEE International Conference on Computer Vision Workshops (ICCVW).

[11]  Andrew Gilman,et al.  Computer-assisted recognition of dolphin individuals using dorsal fin pigmentations , 2016, 2016 International Conference on Image and Vision Computing New Zealand (IVCNZ).

[12]  Mahmoud R. El-Sakka,et al.  North Atlantic right whale localization and recognition using very deep and leaky Neural Network , 2017 .

[13]  Anil Thomas,et al.  NORTH ATLANTIC RIGHT WHALE LOCALIZATION AND RECOGNITION USING VERY DEEP AND LEAKY NEURAL NETWORK , 2017 .

[14]  Ting Dong,et al.  Dolphin fin pose correction using ICP in application to photo-identification , 2013, 2013 28th International Conference on Image and Vision Computing New Zealand (IVCNZ 2013).

[15]  Yu-Wing Tai,et al.  Salient Region Detection via High-Dimensional Color Transform and Local Spatial Support , 2014, IEEE Transactions on Image Processing.

[16]  Pietro Perona,et al.  Graph-Based Visual Saliency , 2006, NIPS.