Portability and Acceleration of Deep Learning Inferences to Detect Rapid Earthquake Damage From VHR Remote Sensing Images Using Intel OpenVINO Toolkit

Abstract

Accurate and effective rapid detection in remote sensing images play an extremely important role in natural disasters, landslides, flooding problems and military defense. Specially in earthquake damage detection, time critical tasks such as performing the damage assessment or providing immediate delivery of relief assistance require responses for swift decisions. To minimize response time, this paper proposes the portability and acceleration of the inferences process on a deep convolution neural network (CNN) model to detect rapid earthquake damage on Very High Resolution (VHR) remote sensing data using the Intel OpenVINO toolkit. This model, based on a previous work, has been optimized for High Performance Computing techniques to minimize processing times. Along with the immediate responses, the correctness of the result is also very crucial parameter for such type of applications. For analysis and performance, we use the Geoeye-1 VHR disaster images of the Haiti earthquake occurred in year 2010. Experimental results show that the optimized model provides good accuracy for damage detection with significant execution speed on CPU+GPU using mixed precision technique compared to other previous work. Moreover, OpenVINO toolkit presents satisfactory performance in the inference stage compared to other toolkits like TensorFlow serving using asynchronous inference mode executed on an Intel Xeon processor and Intel Movidius NCS as co-processor.

Accurate and Fast Classification of Natural Disasters using CNN-LSTM and Inference Acceleration

“…The number of images processed in a single second is known as throughput (FPS), indicating the processing speed of the entire process. The time taken to perform an inference for every input is known as latency [39]. Table VII presents the performance of the pure-CNN, CNN-RAP, and CNN-LSTM models, using the test dataset that is compiled into video format.…”

An Optimized Multi-Task Learning Model for Disaster Classification and Victim Detection in Federated Learning Environments

“…The correctness of the post-training optimization results, especially for model accuracy, is very crucial for actual deployment. The majority of the research in [23][24][25][26] tries to accelerate the inference process without detailing the degree of accuracy loss. In contrast, our measurement outputs are based on open-source and production-ready frameworks to ensure reusability, interoperability, and scalability.…”

“…By recognizing the importance of inference optimization, a plethora of works utilized OpenVINO on various use cases such as license plate detection [23], person re-identification system [24] and face recognition [25]. Work that explicitly optimizes OpenVINO model for disaster scenario was found in [26]. However, all these research tries to accelerate the inference process without detailing the degree of accuracy loss.…”

Hybrid deep learning model with enhanced sunflower optimization for flood and earthquake detection

“…One popular approach for earthquake detection and identification is the short-time average (STA or LTA) [ 10 ]. Nevertheless, STA and LTA have a number of limitations, such as the fact that erroneous initialization values result in a number of false alarms and that they do not work well when a great deal of noise is present [ 11 ].…”