Accurate bee counting is important for colony monitoring, pollination assessment, and precision beekeeping, but manual counting and dense point annotation are labor-intensive. This study proposes M3DANet, a lightweight semi-supervised density regression network with a handheld edge deployment system for bee colony counting. A dataset containing 586 valid high-resolution images and 34,869 point annotations was constructed for training and evaluation. M3DANet uses the first seven stages of MobileNetV3-Large as the lightweight backbone and combines multi-scale context encoding, attention-guided low-level feature fusion, and teacher–student consistency learning with confidence masking and warm-up training. The 10%, 30%, and 50% labeled data settings refer to the proportions of labeled images in the training set, and the remaining training images are used as unlabeled data. Mean absolute error (MAE) and root mean square error (RMSE) are used as evaluation metrics. On the main dataset, M3DANet achieved MAE values of 9.937, 7.003, and 5.570 and RMSE values of 13.093, 9.387, and 7.620 under the 10%, 30%, and 50% settings, respectively, outperforming representative semi-supervised baselines. Under the fully supervised setting, it achieved an MAE of 5.201 and an RMSE of 6.989 with only 2.095 M parameters and 416.64 FPS, using 87.1% fewer parameters and running 17.7 times faster than CSRNet. Cross-species experiments confirmed its low-label generalization ability. Jetson Orin NX deployment achieved 65.75 ms/image inference latency and 10.44 FPS complete-pipeline throughput. These results show that M3DANet balances counting accuracy, annotation efficiency, generalization, and edge deployment practicality.

The power-shift system employed in agricultural multi-range continuously variable transmissions (CVTs) features a complex structure and control logic, presenting significant challenges to the reliability of agricultural machinery. To enable timely detection of faults, constructing an intelligent fault diagnosis classifier to monitor the system’s health status is essential. Typically, fault samples utilized for classifier development originate from ideal bench tests, characterized by uniform patterns and limited diversity, thereby hindering the algorithm’s generalization capability. This study addresses this issue by proposing a generative adversarial network (GAN) model, integrated with a triple loss function and a novel generator architecture, to augment the fault dataset under laboratory conditions. The generator architecture comprises a variational autoencoder module and an oil pressure point attention mechanism, enabling the generation of diverse and fluctuating virtual samples. Building on this augmented dataset, a fault classifier based on one-dimensional ConvNeXt was developed. Experimental results indicate that the classifier achieves an accuracy of 99.73%. While classifier accuracy decreases with increasing noise levels, the GAN-generated dataset provides more comprehensive training, resulting in an accuracy approximately 3% higher than that achieved using the original dataset.

Precise and real-time posture recognition of scallions during harvesting and post-harvest processing is critical for automated conveying, orientation adjustment, bundling, and packaging. Nevertheless, their slender and flexible form, varied spatial orientations, target overlap, lighting fluctuations, background interference, and constrained computational resources of edge devices present significant obstacles to reliable visual perception. This research introduces GOPD-YOLO, a lightweight oriented object-detection network built on the YOLOv8-OBB framework. The network integrates partial-convolution-based lightweight feature extraction to minimize redundant computation, large separable-kernel attention to boost long-range structural representation, and a shared detail-enhanced detection head to improve boundary- and orientation-sensitive prediction. A custom dataset comprising 1500 conveyor-belt images under diverse scallion posture scenarios was developed for model training and assessment. GOPD-YOLO attained a precision of 90.6%, a recall of 94.5%, an mAP@0.5 of 93.2%, and an mAP@0.5:0.95 of 71.5%, with 2.38 million parameters, 6.6 GFLOPs, and a model size of 4.9 MB. Relative to YOLOv8n-OBB, GOPD-YOLO enhanced recall by 3.8 percentage points while decreasing parameter count and model size by 22.7% and 22.2%, respectively. Deployment tests were performed on the Jetson Orin NX Super platform across varying conveyor speeds, lighting conditions, and scallion stacking levels to evaluate the model’s practical utility. These results indicate GOPD-YOLO’s potential as a lightweight vision-based solution for scallion posture recognition in automated harvesting and post-harvest processing.