Hanno Homann

Prof. Dr. Hanno Homann

orcid

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications

5 June 2025 in Applied Sciences
Verified
Journal Article
Open access
PDF file
BHBingqiang Huang
Bingqiang Huang
ZXZongyi Xie
Zongyi Xie
HHHanno Homann
Hanno Homann
ZFZhengshun Fei
Zhengshun Fei
XXXinjian Xiang
Xinjian Xiang
PSPage Ssq
Page Ssq
Detecting small objects in complex outdoor conditions remains challenging. This paper proposes an improved version of YOLOv8n for the detection of blueberry in challenging outdoor scenarios. In this context, this article addresses feature extraction, small-target detection, and multi-scale feature fusion. Specifically, the C2F-RFAConv module is introduced to enhance spatial receptive field learning and a P2-level detection layer is introduced for small and distant targets and fused by a four-head adaptive spatial feature fusion detection head (Detect-FASFF). Additionally, the Focaler-CIoU loss is chosen to mitigate sample imbalance, accelerate convergence, and improve overall model performance. Experiments on our blueberry maturity dataset show that the proposed model outperforms YOLOv8n, achieving 2.8% higher precision, 4% higher recall, and a 4.5% increase in mAP@0.5, with an FPS of 80. It achieves 89.1%, 91.0%, and 85.5% AP for ripe, semi-ripe, and unripe blueberries, demonstrating robustness under varying lighting, occlusion, and distance conditions. Compared to other lightweight networks, the model offers superior accuracy and efficiency. Future work will focus on model compression for real-world deployment.

1 September 2025 in Current Directions in Biomedical Engineering
Verified
Journal Article
Open access
PDF file
SKSofian Khabbich
Sofian Khabbich
JWJens Christian Will
Jens Christian Will
MHMichael Hötter
Michael Hötter
HHHanno Homann
Hanno Homann
Individuals with motor impairments often face challenges in interacting with digital devices. Video-based pose detection has emerged as a promising approach for creating accessible human-machine interfaces (HMIs). However, due to the diverse and highly individualized movement capabilities among this population, generic pose recognition systems are often inadequate. This study investigates the feasibility of a personalized hand-pose recognition system for individuals with motor impairments using minimal training data. We implemented a video-based pose classification model using a webcam and the MediaPipe Hands framework to extract keypoints from hand poses. Each participant, regardless of their level of impairment, selected four individual hand poses according to their capabilities, which were then used to train a neural network. The trained models were evaluated in a realtime test application where participants controlled an animated figure using their hand poses. Results from 14 participants demonstrated that most were able to achieve effective control, even in cases of severe motor limitations. Individuals with restricted finger mobility successfully adapted by utilizing wrist and elbow movements. However, participants with spasticity experienced higher misclassification rates due to difficulties in maintaining stable poses. Overall, our findings highlight the importance of individualized pose recognition systems for assistive technology. Future work should explore additional adaptations, such as head or eye-based controls, to further improve accessibility for users with severe motor impairments.

23 September 2023 in arXiv
Verified
Preprint
Open access
PDF file
HJHenrik Jobczyk
Henrik Jobczyk
HHHanno Homann
Hanno Homann
Generation of computer-aided design (CAD) models from multi-view images may be useful in many practical applications. To date, this problem is usually solved with an intermediate point-cloud reconstruction and involves manual work to create the final CAD models. In this contribution, we present a novel network for an automated reverse engineering task. Our network architecture combines three distinct stages: A convolutional neural network as the encoder stage, a multi-view pooling stage and a transformer-based CAD sequence generator. The model is trained and evaluated on a large number of simulated input images and extensive optimization of model architectures and hyper-parameters is performed. A proof-of-concept is demonstrated by successfully reconstructing a number of valid CAD models from simulated test image data. Various accuracy metrics are calculated and compared to a state-of-the-art point-based network. Finally, a real world test is conducted supplying the network with actual photographs of two three-dimensional test objects. It is shown that some of the capabilities of our network can be transferred to this domain, even though the training exclusively incorporates purely synthetic training data. However to date, the feasible model complexity is still limited to basic shapes.