Using Deep Learning Convolutional Neural System for Discrete Minke Whale Appreciation (Published)
The only known predictable aggregation of dwarf minke whales .occurs in the Australian offshore waters of the northern Great Barrier Reef in May-August each year. The identification of individual whales is required for research on the whales’ population characteristics and for monitoring the potential impacts of tourism activities, including commercial swims with the whales. At present, it is not cost-effective for researchers to manually process and analyze the tens of thousands of underwater images collated after each observation/tourist season, and a large data base of historical non-identified imagery exists. This study reports the first proof of concept for recognizing individual dwarf minke whales using the Deep Learning Convolutional Neural Networks (CNN).The “off-the-shelf” Image net-trained VGG16 CNN was used as the feature-encoder of the per-pixel sematic seg-mentation Automatic Minke Whale Recognizer (AMWR). The most frequently photographed whale in a sample of 76 individual whales (MW1020) was identified in 179 images out of the total 1320 images provided. Training and image augmentation procedures were developed to compensate for the small number of available images. The trained AMWR achieved 93% prediction accuracy on the testing subset of 36 positive/MW1020 and 228 negative/not-MW1020 images, where each negative image contained at least one of the other 75 whales. Furthermore on the test subset, AMWR achieved 74% precision, 80% recall, and 4% false-positive rate, making the presented approach comparable or better to other state-of-the-art individual animal recognition results.
Keywords: convolutional neural networks, deep learning, dwarf minke whales, image recognition, photo-identification, population biology
Analysis on Deep Learning Performance with Low Complexity (Published)
In this article, presenting deep learning to the LTE-A uplink channel estimation system. This work involved creating two SC-FDMA databases, one for training and one for testing, based on three different channel propagation models. The first part of this work consists in applying artificial neural networks to estimate the channel of the SC-FDMA link. Neural network training is an iterative process consisting of adapting the values (weights and biases) of its parameters. After training, the neural network is tested and implemented in the recipient. The second part of this work addresses the same experiment, but uses deep learning, not traditional neural networks. The simulation results show that deep learning has improved significantly compared to traditional methods for bit error rate and processing speed. The third part of this work is devoted to complexity research. Deep learning has been shown to provide better performance than less complex MMSE estimators.
Keywords: ANN, Artificial Intelligence, LTE-A, SC-FDMA, and portable, channel estimation, communications systems, deep learning, mobile