Kristian Muri Knausgård

• Dalal, Anurag; Hagen, Daniel; Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2026). Fast 3D Foundation Model Initialized Gaussian Splatting. 2026 6th International Conference on Electrical, Computer and Energy Technologies (ICECET). doi: 10.1109/icecet65726.2026.11633076. Full text in Research Archive Show summary

  This paper introduces a fast method for high-quality 3D Gaussian Splatting (3DGS) reconstruction without traditional Structure-from-Motion (SfM). The proposed approach leverages 3D Foundation Models (3DFMs) for camera pose and pointcloud initialization, then jointly optimizes both camera poses and Gaussian primitives using a depth-guided loss function. This enables fast convergence even from rough initialization with as few as 50-60 input views. To further improve reconstruction quality in sparse-view scenarios, an MLP-based pose refinement module is introduced alongside depth-guided supervision from the foundation model. Extensive experiments on Mip-NeRF 360, Tanks and Temples, and RobustNeRF demonstrate that the proposed method achieves competitive reconstruction quality (23.61 dB PSNR, 0.19 LPIPS) while reducing training time to approximately three minutes per scene. The proposed method produces ready-to-use 3DGS models at a fraction of the time required by existing pipelines, making it suitable for near realtime applications in robotics, VR, and autonomous navigation.

• Dalal, Anurag; Hagen, Daniel; Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2025). ScatteringSplatting: Capturing Minute Details Using Deep Learning-Aided Gaussian Splatting. International Conference on Control, Mechatronics and Automation. ISSN 2837-5114. p. 547–554. doi: 10.1109/iccma67641.2025.11369605. Full text in Research Archive Show summary

  3D reconstruction in scattering media, such as underwater or foggy environments, presents significant challenges as well as substantial opportunities. Recent advances in novel view synthesis and accurate 3D reconstruction have enabled applications across diverse domains—from ecosystem monitoring and virtual observatories to underwater inspection and exploration using autonomous underwater vehicles and remotely operated vehicles. While Neural Radiance Fields excel at capturing environments with complex lighting effects, their long training times can limit practical applications. In contrast, 3D Gaussian Splatting offers a faster and more explainable alternative for 3D reconstruction and novel view synthesis. The proposed ScatteringSplatting method integrates volumetric rendering with 3D Gaussian Splatting to efficiently reconstruct underwater and foggy scenes. It explicitly separates the scattering media effects, handled via a multi-layer perceptron, from the 3D object geometries processed through Gaussian Splatting, and uses a unique loss function and effective culling of Gaussians to remove unwanted floaters. The primary advantage of our approach becomes evident when generating detailed views, particularly in rendering precise textures and small objects, such as fish. Our method achieves average PSNR, SSIM, and LPIPS scores of 28.91, 0.905, and 0.149, respectively, on the SeaThru-NeRF dataset.

• Dalal, Anurag; Hagen, Daniel; Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2025). Evaluating the effect of image quantity on Gaussian Splatting: A statistical perspective. Computer Vision and Image Understanding. ISSN 1077-3142. 263. doi: 10.1016/j.cviu.2025.104575. Full text in Research Archive Show summary

  3D reconstruction is now a key capability in computer vision. With the advancements in NeRFs and Gaussian Splatting, there is an increasing need on properly capturing data to feed these algorithms and use them in real world scenarios. Most publicly available datasets that can be used for Gaussian Splatting are not suitable to do proper statistical analysis on reducing the number of cameras or the effect of uniformly placed cameras versus randomly placed cameras. The number of cameras in the scene significantly affects the accuracy and resolution of the final 3D reconstruction. Thus, designing a proper data capture system with a certain number of cameras is crucial for 3D reconstruction. In this paper UnrealGaussianStat dataset is introduced, and a statistical analysis is performed on decreasing viewpoints have on Gaussian splatting. It is found that when the number of cameras is increased after 100 the train and test metrics saturates, and does not have significant impact on the reconstruction quality.

• Dalal, Anurag; Hagen, Daniel; Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2024). Gaussian Splatting: 3D Reconstruction and Novel View Synthesis: A Review. IEEE Access. 12, p. 96797–96820. doi: 10.1109/ACCESS.2024.3408318. Full text in Research Archive
• Olsen, Ørjan Langøy; Sørdalen, Tonje Knutsen; Goodwin, Morten; Malde, Ketil; Knausgård, Kristian Muri & Halvorsen, Kim Aleksander Tallaksen (2023). A contrastive learning approach for individual re-identification in a wild fish population. Proceedings of the Northern Lights Deep Learning Workshop. 4. doi: 10.7557/18.6824. Full text in Research Archive Show summary

  In both terrestrial and marine ecology, physical tagging is a frequently used method to study population dynamics and behavior. However, such tagging techniques are increasingly being replaced by individual re-identification using image analysis. This paper introduces a contrastive learning-based model for identifying individuals. The model uses the first parts of the Inception v3 network, supported by a projection head, and we use contrastive learning to find similar or dissimilar image pairs from a collection of uniform photographs. We apply this technique for corkwing wrasse, Symphodus melops, an ecologically and commercially important fish species. Photos are taken during repeated catches of the same individuals from a wild population, where the intervals between individual sightings might range from a few days to several years. Our model achieves a one-shot accuracy of 0.35, a 5-shot accuracy of 0.56, and a 100-shot accuracy of 0.88, on our dataset.

• Holen, Martin; Knausgård, Kristian Muri & Goodwin, Morten (2022). Development of a Simulator for Prototyping Reinforcement Learning-Based Autonomous Cars. Informatics. 9(2). doi: 10.3390/informatics9020033. Full text in Research Archive Show summary

  Autonomous driving is a research field that has received attention in recent years, with increasing applications of reinforcement learning (RL) algorithms. It is impractical to train an autonomous vehicle thoroughly in the physical space, i.e., the so-called ’real world’; therefore, simulators are used in almost all training of autonomous driving algorithms. There are numerous autonomous driving simulators, very few of which are specifically targeted at RL. RL-based cars are challenging due to the variety of reward functions available. There is a lack of simulators addressing many central RL research tasks within autonomous driving, such as scene understanding, localization and mapping, planning and driving policies, and control, which have diverse requirements and goals. It is, therefore, challenging to prototype new RL projects with different simulators, especially when there is a need to examine several reward functions at once. This paper introduces a modified simulator based on the Udacity simulator, made for autonomous cars using RL. It creates reward functions, along with sensors to create a baseline implementation for RL-based vehicles. The modified simulator also resets the vehicle when it gets stuck or is in a non-terminating loop, making it more reliable. Overall, the paper seeks to make the prototyping of new systems simple, with the testing of different RL-based systems.

• Gupta, Aditya; Bringsdal, Even; Salbuvik, Nicole; Knausgård, Kristian Muri & Goodwin, Morten (2022). An Accurate Convolutional Neural Networks Approach to Wound Detection for Farmed Salmon. In Iliadis, Lazaros; Jayne, Chrisina; Tefas, Anastasios & Pimenidis, Elias (Ed.), Engineering Applications of Neural Networks - 23rd International Conference, EAAAI/EANN 2022. Springer Nature. ISSN 9783031082221. p. 139–149. doi: 10.1007/978-3-031-082. Full text in Research Archive Show summary

  Fish is one of the most important food sources worldwide and for people in Nordic countries. For this reason, fish has been widely cultivated, but aquacultural fish are severely affected by lice, maturity, wounds, and other harmful factors typically part of agricultural fish, resulting in millions of fish deaths. Unfortunately, diagnosing injuries and wounds in live salmon fish is difficult. However, this study uses image-based machine learning approaches to present a wound detection technique for live farmed salmon fish. As part of this study, we present a new dataset of 3571 photos of injured and non-wounded fish from the Institute of Marine Research’s genuine fish tank. We also propose a Convolutional Neural Network tailored for such wound detection with 20 convolutional and five subsequent dense layers. The model incorporates methods such as dropout, early halting, and Gaussian noise to avoid overfitting. Compared to the established VGG-16 and VGG-19 models, the proposed approaches have a validation accuracy of 96.22%. The model has low 0.0199 and 0.941 false positive and true positive rates, making it a good candidate for accurate live production

• Holen, Martin; Knausgård, Kristian Muri & Goodwin, Morten (2022). An Evaluation of Autonomous Car Simulators and Their Applicability for Supervised and Reinforcement Learning. In Sanfilippo, Filippo; Granmo, Ole-Christoffer; Yayilgan, Sule Yildirim & Bajwa, Imran Sarwar (Ed.), Intelligent Technologies and Applications: 4th International Conference, INTAP 2021, Grimstad, Norway, October 11–13, 2021, Revised Selected Papers. Springer. ISSN 9783031105241. p. 367–379. doi: 10.1007/978-3-031-10525-8_29. Full text in Research Archive Show summary

  Recent advancements in the field of Machine Learning have sprouted a renewed interest in the area of autonomous cars. Companies use different techniques to develop autonomous cars, including buying several vehicles to develop Advanced Driver Assistance Systems (ADAS), while others use car simulators. Simulators for autonomous cars include a wide variety of different sensors. Some simulators come free or even open-source, while others require to pay for the simulator itself, server time, or each sensor. The quality and focus of each of these vary, with some having LIDAR scanned roads for highly realistic roads, while others have entirely natural and realistic vehicle dynamics.

• Knausgård, Kristian Muri; Skar, Siv; Sanfilippo, Filippo; Buldenko, Albert; Lindheim, Henning & Lunde, Jakob E. [Show all 8 contributors for this article] (2022). Branch-Manoeuvring Capable Pipe Cleaning Robot for Aquaponic Systems. In Sanfilippo, Filippo; Granmo, Ole-Christoffer; Yayilgan, Sule Yildirim & Bajwa, Imran Sarwar (Ed.), Intelligent Technologies and Applications: 4th International Conference, INTAP 2021, Grimstad, Norway, October 11–13, 2021, Revised Selected Papers. Springer. ISSN 9783031105241. p. 107–118. doi: 10.1007/978-3-031-10525-8_9. Full text in Research Archive Show summary

  Aquaponic systems are engineered ecosystems combining aquaculture and plant production. Nutrient rich water is continuously circulating through the system from aquaculture tanks. A biofilter with nitrifying bacteria breaks down fish metabolism ammonia into nitrite and nitrate, which plants and makes the aquaculture wastewater into valued organic fertiliser for the plants, containing essential macro and micro elements. At the same time, the plants are cleaning the water by absorbing ammonia from the fish tanks before it reaches dangerous levels for the aquatic animals. In principle, the only external input is energy, mainly in the form of light and heat, but fish food is also commonly provided. Growing fish food is potentially feasible in a closed loop system, hence aquaponic systems can possibly be an important source of proteins and other important nutrition when, for example, colonising other planets in the future. Fully autonomous aquaponic systems are currently not available. This work aims at minimising manual labour related to cleaning pipes for water transport. The cleaning process must be friendly to both plants and aquatic animals. Hence, in this work, pure mechanical cleaning is adopted. A novel belt-driven continuum robot capable of travelling through small/medium diameter pipes and manoeuvring branches and bends, is designed and tested. The robot is modular and can be extended with different cleaning modules through an interface providing CAN-bus network and electric power. The flexible continuum modules of the robot are characterised. Experimental results demonstrate that the robot is able to travel through pipes with diameters varying from 50 mm to 75 mm, and also capable of handling T-branches of up to 90∘.

• Gupta, Aditya; Bringsdal, Even; Knausgård, Kristian Muri & Goodwin, Morten (2022). Accurate Wound and Lice Detection in Atlantic Salmon Fish Using a Convolutional Neural Network. Fishes. 7(6). doi: 10.3390/fishes7060345. Full text in Research Archive Show summary

  The population living in the coastal region relies heavily on fish as a food source due to their vast availability and low cost. This need has given rise to fish farming. Fish farmers and the fishing industry face serious challenges such as lice in the aquaculture ecosystem, wounds due to injuries, early fish maturity, etc. causing millions of fish deaths in the fish aquaculture ecosystem. Several measures, such as cleaner fish and anti-parasite drugs, are utilized to reduce sea lice, but getting rid of them entirely is challenging. This study proposed an image-based machine-learning technique to detect wounds and the presence of lice in the live salmon fish farm ecosystem. A new equally distributed dataset contains fish affected by lice and wounds and healthy fish collected from the fish tanks installed at the Institute of Marine Research, Bergen, Norway. A convolutional neural network is proposed for fish lice and wound detection consisting of 15 convolutional and 5 dense layers. The proposed methodology has a test accuracy of 96.7% compared with established VGG-19 and VGG-16 models, with accuracies of 91.2% and 92.8%, respectively. The model has a low false and true positive rate of 0.011 and 0.956, and 0.0307 and 0.965 for fish having lice and wounds, respectively.

• Goodwin, Morten; Halvorsen, Kim Aleksander Tallaksen; Jiao, Lei; Knausgård, Kristian Muri; Martin, Angela Helen & Moyano, Marta [Show all 10 contributors for this article] (2022). Unlocking the potential of deep learning for marine ecology: overview, applications, and outlook. ICES Journal of Marine Science. ISSN 1054-3139. 79(2), p. 319–336. doi: 10.1093/icesjms/fsab255. Full text in Research Archive Show summary

  The deep learning (DL) revolution is touching all scientific disciplines and corners of our lives as a means of harnessing the power of big data. Marine ecology is no exception. New methods provide analysis of data from sensors, cameras, and acoustic recorders, even in real time, in ways that are reproducible and rapid. Off-the-shelf algorithms find, count, and classify species from digital images or video and detect cryptic patterns in noisy data. These endeavours require collaboration across ecological and data science disciplines, which can be challenging to initiate. To promote the use of DL towards ecosystem-based management of the sea, this paper aims to bridge the gap between marine ecologists and computer scientists. We provide insight into popular DL approaches for ecological data analysis, focusing on supervised learning techniques with deep neural networks, and illustrate challenges and opportunities through established and emerging applications of DL to marine ecology. We present case studies on plankton, fish, marine mammals, pollution, and nutrient cycling that involve object detection, classification, tracking, and segmentation of visualized data. We conclude with a broad outlook of the field’s opportunities and challenges, including potential technological advances and issues with managing complex data sets.

• Knausgård, Kristian Muri; Wiklund, Arne; Sørdalen, Tonje Knutsen; Halvorsen, Kim Tallaksen; Kleiven, Alf Ring & Lei, Jiao [Show all 7 contributors for this article] (2021). Temperate fish detection and classification: a deep learning based approach. Applied Intelligence - The International Journal of Research on Intelligent Systems for Real Life Complex Problems. ISSN 0924-669X. 52. doi: 10.1007/s10489-020-02154-9. Full text in Research Archive Show summary

  A wide range of applications in marine ecology extensively uses underwater cameras. Still, to efficiently process the vast amount of data generated, we need to develop tools that can automatically detect and recognize species captured on film. Classifying fish species from videos and images in natural environments can be challenging because of noise and variation in illumination and the surrounding habitat. In this paper, we propose a two-step deep learning approach for the detection and classification of temperate fishes without pre-filtering. The first step is to detect each single fish in an image, independent of species and sex. For this purpose, we employ the You Only Look Once (YOLO) object detection technique. In the second step, we adopt a Convolutional Neural Network (CNN) with the Squeeze-and-Excitation (SE) architecture for classifying each fish in the image without pre-filtering. We apply transfer learning to overcome the limited training samples of temperate fishes and to improve the accuracy of the classification. This is done by training the object detection model with ImageNet and the fish classifier via a public dataset (Fish4Knowledge), whereupon both the object detection and classifier are updated with temperate fishes of interest. The weights obtained from pre-training are applied to post-training as a priori. Our solution achieves the state-of-the-art accuracy of 99.27% using the pre-training model. The accuracies using the post-training model are also high; 83.68% and 87.74% with and without image augmentation, respectively. This strongly indicates that the solution is viable with a more extensive dataset.

• Olsvik, Erlend; Trinh, Christian M.D.; Knausgård, Kristian Muri; Wiklund, Arne; Sørdalen, Tonje Knutsen & Kleiven, Alf Ring [Show all 8 contributors for this article] (2019). Biometric fish classification of temperate species using convolutional neural network with squeeze-and-excitation. Lecture Notes in Computer Science (LNCS). ISSN 0302-9743. 11606 LNAI, p. 89–101. doi: 10.1007/978-3-030-22999-3_9. Full text in Research Archive

View all works in NVA

• Holen, Martin; Ruud, Else-Line Malene; Warakagoda, Narada Dilp; Granmo, Ole-Christoffer; Engelstad, Paal E. & Knausgård, Kristian Muri (2022). Towards Using Reinforcement Learning for Autonomous Docking of Unmanned Surface Vehicles. doi: 10.1007/978-3-031-08223-8_38. Full text in Research Archive Show summary

  Providing full autonomy to Unmanned Surface Vehicles (USV) is a challenging goal to achieve. Autonomous docking is a subtask that is particularly difficult. The vessel has to distinguish between obstacles and the dock, and the obstacles can be either static or moving. This paper developed a simulator using Reinforcement Learning (RL) to approach the problem. We studied several scenarios for the task of docking a USV in a simulator environment. The scenarios were defined with different sensor inputs and start-stop procedures but a simple shared reward function. The results show that the system solved the task when the IMU (Inertial Measurement Unit) and GNSS (Global Navigation Satellite System) sensors were used to estimate the state, despite the simplicity of the reward function.

• Knausgård, Kristian Muri; Skar, Siv; Sanfilippo, Filippo; Buldenko, Albert; Lindheim, Henning & Lunde, Jakob [Show all 8 contributors for this article] (2021). Branch-Manoeuvring Capable Pipe Cleaning Robot for Aquaponic Systems. Full text in Research Archive Show summary

  Aquaponic systems are engineered ecosystems combining aquaculture and plant production. Nutrient rich water is continuously circulating through the system from aquaculture tanks. A biofilter with nitrifying bacteria breaks down fish metabolism ammonia into nitrite and nitrate, which plants and makes the aquaculture wastewater into valued organic fertiliser for the plants, contenting essential macro and micro elements. At the same time, plants are cleaning the water by absorbing ammonia from the fish tanks before it reaches dangerous levels for the aquatic animals. In principle, the only external input is energy, mainly in the form of light and heat, but fish food is also commonly provided. Growing fish food is potentially feasible in a closed loop system, hence aquaponic systems can possibly be an important source of proteins and other important nutrition when, for example, colonising other planets in the future. Fully autonomous aquaponic systems are currently not available. This work aims at minimising manual labour related to cleaning pipes for water transport. The cleaning process must be friendly to both plants and aquatic animals. Hence, in this work, pure mechanical cleaning is adopted. A novel belt-driven continuum robot capable of travelling through small/medium diameter pipes and manoeuvring branches and bends, is designed and tested. The robot is modular and can be extended with different cleaning modules through an interface providing CAN-bus network and electric power. The flexible continuum modules of the robot are characterised. Experimental results demonstrate that the robot is able to travel through pipes of diameter 50 mm to 75 mm, and also capable of handling T-branches of up to 90 degrees.

• Robbersmyr, Kjell Gunnar; Choux, Martin Marie Hubert; Zhou, Jing & Knausgård, Kristian Muri (2018). Mekatronikk. Full text in Research Archive
• Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2018). Mekatronikk. Full text in Research Archive
• Knausgård, Kristian Muri & Ottestad, Morten (2018). Mobile autonome roboter i et utdanningsperspektiv. Full text in Research Archive
• Robbersmyr, Kjell Gunnar & Knausgård, Kristian Muri (2018). Studentprosjekter 2018. Full text in Research Archive
• Gaustad, Eskil Gresen; Knausgård, Kristian Muri & Zhou, Jing (2025). Supervisory Control Design for a SES Vessel. Universitetet i Agder. Full text in Research Archive
• Knausgård, Kristian Muri & Robbersmyr, Kjell Gunnar (2018). Studentprosjekter - innhenting og valg. Universitetet i Agder. Full text in Research Archive
• Esnault, Michel; Knausgård, Kristian Muri & Robbersmyr, Kjell Gunnar (2017). Innovation Gateway – Behovskartlegging og struktur. Universitetet i Agder. Full text in Research Archive

View all works in NVA