# DeepShip

keeping in view the limitation of file size and space on github directory, a part of dataset could be uploaded on this repository. The remaining part can be downloaded by 
sending the email to [email protected] 

DeepShip offers an opportunity to train and evaluate the performance of different algorithms, and to identify their strengths and weakness. The advantage of this dataset is that
it is recorded in the real world sea environment in different seasons and sea conditions. Along with vessels signals, recorded signals also have natural background noise, marine
mammal noise and noise of any other human initiated activity. This dataset consists of 47 hours and 04 minutes of real world underwater recordings of 265 different ships belong
to four classes. Data for four commercial ship classes is provided. These classes include oil tanker, tug, passenger ship and cargo ship. Moreover,
mechanism of collection of data and its labeling are also discussed. We hope that availability of a good benchmark data set will promote and accelerate the research and
development in this field. 
Data recording is done in duration between 02 May 2016 to 04 October 2018 in strait of Georgia delta node. Figure 2 illustrates the over all setup for recording of vessel audio.
Data recording site is situated in one of the busiest shipping routes of the Pacific Northwest coast, and near to the busiest port in Canada, i.e. Vancouver. The background
has great influence of river discharge and strong tidal currents with presence of vibrant biological marine environment. The seabed composition in this area is comprised
of silt and sand sediment. The central location of strait of Georgia is strongly influenced by semi-diurnal tidal currents, which vary over the spring-neap cycle with 
speeds ranging between 1 and 3 knots (0.5 - 1.5 m/s). Further, in the summer and early fall, deep dense gravity currents periodically sweep along the bottom. In this area 
main species include cetaceans (whales and dolphins) and salmon. The southern area is acutely influenced by human activity such as recreational boating, shipping, and 
industry affecting the acoustic environment of the ocean. Based on above mentioned reasons it is to mention that the recorded signals may contain background noise both 
from human and marine life activities.

The data is acquired by using an ocean sonics icListen AF hydrophone, as shown in Figure 3. Main specifications of the used hydrophone are shown in Table 2. The icListen 
smart hydrophone is a broadband digital ultra-quit hydrophone, with bandwidth 1 Hz-12 kHz, dynamic range 120 dB, sensitivity -170 dBV re. � Pa, power&input 12-24 Vdc, 0.8 W, 
case material is engineered plastic and titanium with depth capacity of 200m & 3500m respectively. It is a compact, all-in-one instrument designed and manufactured by 
replacing separate pre-amplifier, filters, converters and data link units with a compact unit, capable of processing real time acquired data with direct digital output.