Efficient deep learning in space: Knowledge distillation and optimization of resource usage in a satellite

dc.contributor.authorAsaad, Ebaa
dc.contributor.authorLarsson, Sara
dc.contributor.departmentChalmers tekniska högskola / Institutionen för mekanik och maritima vetenskapersv
dc.contributor.examinerBenderius, Ola
dc.contributor.supervisorBenderius, Ola
dc.contributor.supervisorAnlind, Alice
dc.contributor.supervisorvon Essen, Hannes
dc.contributor.supervisorTranheden, Wilhelm
dc.date.accessioned2022-07-07T11:42:38Z
dc.date.available2022-07-07T11:42:38Z
dc.date.issued2022sv
dc.date.submitted2020
dc.description.abstractThe development of micro-satellites and machine learning (ML) has increased drastically in recent years, which has unlocked new possibilities in the field of Earth observation. One of the applications is the tracking of maritime vessels since the current tracking systems such as the automatic identification system (AIS) can be eluded by simply switching it off. This project, therefore, investigates the possibilities of applying ML in a satellite, specifically with the aim of detecting maritime vessels. An object detector (YOLOv5) was chosen for testing due to its speed, small size, and its user-friendly framework. For comparison with a simpler model, the classification models ShuffleNetV1 and a custom-built CNN model were chosen. Thereafter, for the purpose of optimization, knowledge distillation, as well as different methods for reducing resource usage, were tested. The results show that it is feasible to implement ML on board a satellite to detect maritime vessels, where the best result for YOLOv5 was 2.1 min per 10,000×10,000 pixels RGB image on the target hardware using the GPU. Using the CPU with multiple threads achieved a result of 2.2 min for the same image. Increasing the batch size did not yield better results. ShuffleNetV1 was not supported by the TFLite framework, due to a network structure called group convolutions. Neither was quantization supported by the target hardware, but it did decrease the file size of the model by half. Using knowledge distillation showed great results for the classifiers. Using ShuffleNetV1 to train the simpler CNN model yielded an increase of 12 % in accuracy. It also shows that it is possible to apply a non-supported network on the target device by distilling the knowledge to a supported network. Distilling knowledge using YOLOv5 was more difficult, due to the complexity of the network and the task of object detection. Two methods were therefore tested: using the teacher’s output (logits) as a target and using the feature maps within the network as targets (feature imitation). However, only minimal improvements were reflected in the results.sv
dc.identifier.coursecodeMMSX30sv
dc.identifier.urihttps://hdl.handle.net/20.500.12380/305120
dc.language.isoengsv
dc.relation.ispartofseries2022:33sv
dc.setspec.uppsokTechnology
dc.subjectsatellitessv
dc.subjectEarth observationsv
dc.subjectspacesv
dc.subjectmaritime vesselssv
dc.subjectYOLOv5sv
dc.subjectShuffleNetV1sv
dc.subjectTensorflow Litesv
dc.subjectknowledge distillationsv
dc.subjectoptimizationsv
dc.subjectneural networkssv
dc.titleEfficient deep learning in space: Knowledge distillation and optimization of resource usage in a satellitesv
dc.type.degreeExamensarbete för masterexamensv
dc.type.uppsokH
Ladda ner
Original bundle
Visar 1 - 1 av 1
Bild (thumbnail)
Namn:
2022-33 Ebaa Asaad & Sara Larsson.pdf
Storlek:
3.94 MB
Format:
Adobe Portable Document Format
Beskrivning:
License bundle
Visar 1 - 1 av 1
Bild saknas
Namn:
license.txt
Storlek:
1.51 KB
Format:
Item-specific license agreed upon to submission
Beskrivning: