Deep learning-based morphometric analysis of zebrafish is widely utilized for non-destructively identifying abnormalities and diagnosing diseases. However, obtaining discriminative and continuous organ category decision boundaries poses a significant challenge by directly observing zebrafish larvae from the outside. To address this issue, this study simplifies the organ areas to polygons and focuses solely on the endpoint positioning. Specifically, we introduce a deep learning-based feature endpoint detection method for quantitatively determining zebrafish larvae’s essential phenotype and organ features. We propose the cross-shaped heat tensor network (CSHT-Net), a feature point detection framework consisting of a novel keypoint training method named cross-shaped heat tensor and a feature extractor called combinatorial convolutional block. Our model alleviates the problem of the heatmap-based method that restricts attention to local regions around key points while enhancing the model’s ability to learn continuous, strip-like features. Moreover, we compiled a dataset of 4389 bright-field micrographs of zebrafish larvae at 120 h post-fertilization for the model training and algorithm evaluation of zebrafish phenotypic traits. The proposed framework achieves an average precision (AP) of 83.2% and an average recall (AR) of 85.8%, outperforming multiple widely adopted keypoint detection techniques. This approach enables robust phenotype extraction and reliable morphometric analysis for zebrafish larvae, fostering efficient hazard identification for chemicals and medical products.

Large-scale high-resolution (HR) land cover mapping is essential in monitoring the Earth’s surface and addressing critical challenges facing humanity. While weakly supervised methods help to mitigate the scarcity of HR annotations across wide geographic areas, existing approaches struggle with feature extraction instability. To address this issue, this study proposes AECA-FBMamba, an efficient weakly supervised framework that enhances model perception by stabilizing feature transitions during encoding. Specifically, this work introduces the Adaptive Environment Channel Alignment (AECA) module at the input stage, processing independently grouped color channels to enhance robust channel-wise feature extraction. Additionally, we incorporate the Feature Bridging Mamba (FBMamba) module, which enables smooth receptive field reduction, effectively addressing feature alignment issues when integrating local contexts into global representations. The proposed AECA-FBMamba achieved a 65.27% mIoU on the Chesapeake Bay dataset and a 56.96% mIoU on the Poland dataset. Experiments conducted on these two large-scale datasets demonstrate the method’s effectiveness in automatically updating high-resolution (HR) land cover maps using low-resolution (LR) historical annotations. This framework advances weakly supervised learning in remote sensing and offers solutions for large-scale land cover mapping applications.